Sulfur Gases and Aerosols in and Above the Equatorial African Rain Forest

,

the wet season (ABLE 2B), even lower soil emission values were found at the same site [Andreae et al., 1990a].The total emission of reduced sulfur gases from soils and the plant canopy was estimated to be about 2 nmol m -2 min -• with little variation between seasons.Sulfur gas concentrations over Amazonia were about 10-40 ppt for DMS and H2S in the canopy, decreasing to about 1-2 ppt for DMS in the free troposphere.CH3SH was an order of magnitude lower.
From the ABLE 2 data an internally consistent picture was established between low fluxes of the short-lived sulfur gases from the soil/canopy system to the atmosphere, their fluxes in the atmosphere above the canopy, their concentration and transformation in the atmosphere, the reaction products, and their removal from the atmosphere by dry and wet deposition [Andreae and Andreae, 1988;Andreae et al., 1990aAndreae et al., , 1990b]].This picture, if representative of the cycling of sulfur over remote continental ecosystems, indicates total fluxes of biogenic short-lived sulfur compounds to the atmosphere of about 0.15 Gmol S yr -• , which falls at the very low end of existing global sulfur budgets [Freney et al., 1983].In view of the large discrepancy between sulfur fluxes reported from tropical Africa and those from Amazonia, it was considered important to conduct measurements in Africa using the same methods that had been applied in Amazonia, in order to make possible a comparison of the cycling of sulfur compounds in the wet tropical ecosystems of South America and Africa.Here, we present the results of measurements made in and above the African equatorial rain forest during the 1988 Dynamique et Chimie de l'Atmosphb, re en For•,t Equatoriale (DECAFE) experiment.

Samples for the determination of H2S, DMS, and COS
were collected aboard a small aircraft from the boundary layer and the free troposphere.Air was sampled from an FEP Teflon manifold which was continuously flushed with outside air flowing in through Teflon tubing (8 mm ID) at a flow rate of approximately 15 L min-•.The inlet was located about 15 cm upstream of the nose of the aircraft and brought to the manifold by about 3 m of Teflon tubing.Sample volumes were measured by integrating mass flowmeters (Teledyne-Hastings-Raydist).COS was sampled by filling evacuated stainless steel canisters.At ground level, samples were collected at various heights within the forest and a clearing nearby.Here, battery-powered pumps were used for drawing ambient air through the sampling devices, and gas sample volumes were determined from flow rates measured with rotameters and from sampling time.Throughout this paper, mixing ratios are reported in parts per thousand on a molar basis, whereby 1 ppt equals 10 -•2 mol of a species per mol of air.
Hydrogen sulfide was sampled and analyzed according to the method of Natusch et al. [1972], with the modifications recently developed by Saltzman and Cooper [1988] to overcome a positive interference created by atmospheric COS.For preconcentration of H2S, air was drawn at a rate of 7-11 L min-• through silver nitrate impregnated filters (Whatman 41, 47 mm diameter) which were held in PFE Teflon filter holders.Sulfide collected on the filters was recovered by rinsing with 20 ml NaOH/NaCN solution and then determined fluorometrically by fluorescence quenching of dilute fluorescein mercuric acetate.We used a Turner Designs model l0 series fluorometer with a blue lamp and l0 nm bandpass filters at 500 nm excitation and 520 nm emission wavelengths.Calibration in the field was performed using freshly prepared standard solutions of sodium sulfide [Natusch et al., 1972].After the expedition, in the laboratory, further calibration studies were made using solutions of anhydrous sodium sulfide (Alfa Research Chemicals, Karlsruhe, West Germany, Catalog 65122) and certified H2S permeation devices (VICI Metronics, Santa Clara, California).The permeation devices were stored at 30øC and permeation rates were determined gravimetrically.They were placed in a temperature-controlled dynamic dilution system to produce gas streams with 50-300 ppt H2S.Comparison between calibrations based on the dissolved sulfide standards and those based on the gas dilution system showed that the recovery from the gas phase was 89%.The field data from DECAFE 88 were corrected by this factor.Atmospheric COS presents a positive interference of the method, since approximately 1% of it hydrolyzes on the filters.It is recovered during analysis and thus produces an artifact (typically about $-10 ppt) H2S signal [Saltztnan and Cooper, 1988].To overcome this interference, two filters were used in tandem during sampling.The backup filters were used to obtain a measure of the COS artifact, which was then subtracted from the signal on the front filter, yielding a corrected H2S value.The detection limit of the method (signal to noise of 2) is 13 ppt at 500 L sample volume.Under field conditions in Africa the precision at this level is 14%.All samples were analyzed in a field laboratory within 3-6 hours after sampling.
DMS was sampled by adsorption onto gold wool held in quartz glass tubes.Total sample volumes of 50-100 L were collected at a rate of up to 2 L min -• .After sampling, the tubes were capped and stored in sealed containers.The samples were shipped back to the laboratory after the campaign, where DMS was desorbed thermally from the gold and analyzed by 6C/FPD.This technique has been described in detail previously [Andreae et al., 1985].As a check, 24 blank tubes were stored together with the samples, and no significant contamination during storage was found.The analytical system was calibrated from a dilution system in which nitrogen was passed over a DMS permeation tube held at constant temperature.Samples of this gas stream were collected onto gold tubes and measured.The detection limit of the method is approximately 1.5 ppt DMS at a sample volume of 40 L. Precision and accuracy are typically near __ 10 and __20%, respectively, at 50 ppt DMS.
Atmospheric COS was sampled by filling evacuated 6-L electropolished stainless steel canisters with ambient air.The canister samples were shipped to France and analyzed in the CNRS laboratory.For analysis, COS was cryotrapped at -120øC on TENAX-6C from aliquots of the canister samples, desorbed thermally, separated on a Varian 3400 6C, and detected with a dual-flame FPD [Belviso et al., 1987].The samples were analyzed less than 1 month after they had been collected in the field.
To check for changes in sample composition due to storage, test samples of natural air were stored in the containers and analyzed over a period of 1 month.No significant variations of COS were observed.The precision of the analysis of COS in air samples is about 7% and the detection limit is 0.4 ng of COS.The reproducibility is better than 5%.
Aerosol was sampled at the forest site in the canopy about 5 m above the ground.Stacked filter units (SFU) were used to size-segregate coarse (2.0-15 /•m) and fine (<2.0 /•m) particles according to the procedure described by Artaxo et al. [1988].Samples were loaded for 2 days at a flow rate of approximately 14 L min -1.The SFU were stored in sealed containers before and after loading.Blank filters received the same handling as exposed filters before and after sampling.The samples were analyzed for 19 elements (A1, Si, P, S, C1, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Br, Sr, Rb, and Pb) by particle-induced X ray emission (PIXE).A 2.4 MeV proton beam of the Van-der-Graaf accelerator from PUC University in Rio de Janeiro, Brazil, was used.The detection limit for elemental sulfur and potassium reported here was typically 5 ng m-3 and the precision of the PIXE analysis for these elements was better than 5%.
A single rainfall event was sampled at Improndo.After a dry period of at least 10 days a thunderstorm brought heavy showers at around midnight on February 22. Rainwater was sampled with a polyethylene bucket in front of the field laboratory, at about 40 cm above ground.The bucket had been rinsed with deionized water and was exposed a few minutes before the rain began.The samples were stored in polyethylene bottles with a small amount of chloroform added to suppress microbial activity.The samples were analyzed in Mainz for sulfate using ion chromatography, with a DIONEX HPIC-AS4 column and 2.1 mM NaHCO3/ 1.9 mM Na2CO3 eluent.

Sampling Environment and Meteorological Situation
The

Emissions of H2S and DMS from the Congo Rain Forest
The vertical decay of H2S and DMS and the absence of pronounced maxima in the profiles, as was observed at around 2-2.5 km for primary and secondary burning products like CO, C2H2, 03, and aerosols, reflects the dominance of biogenic emissions by the plant/canopy system for these gases as well as their short lifetimes.The biogenic emission fluxes of H2S and DMS can be estimated by solving the continuity equation for the atmospheric concentrations of H2S and DMS in the boundary layer over the forest and constraining the solutions to match the atmospheric concentrations observed from aircraft (Figures 1 and 3).We view the boundary layer over the forest as horizontally well mixed, which seems a reasonable assumption considering the homogeneity of the terrain and the undisturbed meteorological conditions observed throughout the experiment.(Table 1), whereas the aerosol sulfur concentrations measured during the experiment averaged 248 ppt.We conclude that biogenic sulfur emissions from the forest made only a small contribution to the aerosol sulfur budget.[Andreae, 1983] and possibly from live plants [Beaufort et al., 1977], may serve as an indicator of pyrogenic and biogenic contributions to atmospheric aerosol.Fine mode sulfur and potassium present in excess of their terrigenic components were derived by subtracting soil-derived potassium and sulfur from the observed concentrations (Table 2), using iron as a tracer for the soil component together with Bowen's composition of crustal rock [Bowen, 1979]

CONCLUSIONS
The vertical profiles of the short-lived sulfur species H2S and DMS from the ground level up to the free troposphere are both in their structure and numerical values remarkably similar to those found over Amazonia during the dry season.Thus no evidence of drastic differences in the cycling of biogenic sulfur between the equatorial forest ecosystems of South America and those of Africa could be demonstrated.
Airborne particulate sulfur and rainwater sulfate were found to be slightly higher than during the Amazon

Fig. 1 .Fig. 2 .Fig. 3 .
Fig. 1.Vertical distribution of H2S, obtained during DECAFE 88: (upper part) from sampling on-board aircraft and (lower part) at 0.5, 10, and 20 m in the forest canopy.Error bars given represent analytical and sampling uncertainty (1 sin).Solid circles represent the results of the model simulation with 0.4 ppb NOx.
assumed as upper boundary conditions at the top of the boundary layer.Lower boundary conditions are defined by the emission fluxes cI)(t) of H2S and DMS at canopy top, which vary with time of day to reflect the dependence of vegetative emissions on sunlight and temperature [Fall et al., 1988].We express this observations indicate a welldefined diurnal cycle of mixed layer growth and decay, with dry mixing depths (zi) peaking at =600 m in early afternoon.The dry mixed layer was capped in the daytime by a cloud convective layer (CCL), which carried turbulent energy from the mixed layer aloft in shallow convective cells.The CCL grew in phase with the mixed layer, with CCL depths (Zc) peaking at =2000 m in early afternoon.The top of the CCL at z•. was marked by a strong persistent inversion.Midday values of z i and •. were highly reproducible from day to day over the course of the experiment (R. Lyra et al.(cm s-l) is the convective velocity scale.The term aw*zi is added in (3a) for continuity of K at zi.Values of zi,

ZcFig. 4 .
Fig. 4. Simulated vertical distribution of OH (noon) and NO 3 (midnight) in the atmosphere over the Congo rain forest.

Fig. 5 .
Fig. 5. Concentration of sulfur and potassium in fine and coarse mode aerosol sampled in the rain forest canopy at 5 m height.

Fig. 7 .
Fig. 6.Vertical distribution of COS during DECAFE 88 (composite of data from several flights and ground-based samples).
's dry season, most likely due to a substantial contribution to the abundance of the longer-lived sulfur species SO• and SO2 by burning activities in the vast arid regions north of equatorial Africa.This might explain some of the discrepancies which currently exist between the high sulfur deposition fluxes reported from equatorial Africa [Delrnas and Servant, 1983] and recent estimates for equatorial South America [Andreae and Andreae, 1988; Andreae et al., 1990a].On the basis of the vertical distribution of COS and its relationship to other gases derived from burning, a substantial contribution of biomass burning to the abundance of atmospheric COS appears likely.Our data set shows no obvious evidence for uptake of COS through the plant canopy, probably because COS emissions from fires mask any effect of sinks on the vertical profile.

TABLE 2 . Sulfur and Potassium (ppt) in Fine Mode Aerosol From the Rain Forest Canopy During DECAFE 88
long-