A paper by Charlson et. al, (1987), adopted the working hypothesis that biogenic sulfur might have evolved to stabilize climate in a feedback thermostatic loop employing the Twomey "indirect" aerosol effect. That paper suggested that biogenic sulfur may undergo variation in response to biologically-sensitive climate parameters (e. g. light, sea temperature, etc.). Shaw (1983) suggested a similar mechanism, in which biogenic sulfate aerosol increases the planetary albedo to offset increasing sea surface temeprataures and ocean sulfur emissions. Either of these "biological" mechanisms could stabilize climate, but both would require "active" participation by the biota, such as the biologically-induced increasing quantities of di-methyl sylfide or its parent in response to increased sunlight or ocean temperature. Ramanathan and Collins (1991), noting that sea surface temperature always remains below 302K, suggested an alternative thermostatic feedback loop, having nothing to do with sulfate aerosols or biology, but cooling the tropical climate by increasing albedo from cirrus blowing off the top of deep convective cumulonimbus systems. That paper did not resolve, and neither did subsequent experiments, if the higher albedo would introduce sufficient cooling to offset both the increases in sea surface temperatures and the cloud greenhouse forcing (which becomes substantial for the high altitude Ci).
In the present paper, we suggest an alternative climate stabilizing feedback loop that incorporates biogenic sulfur pumped from the seas into the middle and high troposphere by widescale deep convection. Aerosols nucleating and growing from this parent biogenic sulfur free tropospheric source subside into the marine boundary layer (e. g. Raes, 1995; Raes and Dingeneen, 1992, Raes et. al, 1993). After subsiding into the marine boundary layer we propose that they play an important role in nucleating clouds, thereby reducing absorption of solar energy by the oceans. An increase in sea surface temperature would lead to more convection ( e. g. Chatfield and Crutzen, 1984) , more aerosols,higher cloud drooplet numbers, higher albedo of marine stratus, reduced absorption of solar radiation by oceans, reduced sea surface temperature, reduced convection, thereby providing a potential for regulation of earth climate. This is schematically illustrated by the drawing in Figure 1.
This proposed feedback loop shown in Figure 1 has two important strategic components:
1) it incorporates biogenic sulfur in an incidental manner, not requiring the biological systems to participate actively. We hypothesize that the magnitude of the feedback is controlled by convective transport of sulfur to the upper troposphere. In this sense, the thermostating mechanism and the biota may have co-evolved, a concept elaborated by Schneider and Londer (1984).
2) Cooling near the surface is ensured by the proposed loop since an increase in sulfur flux by convection can only lead to brightening of stratocumulus and cumulus clouds in the marine boundary layer. No ambiguity arises about the sign of the response of the climate system to increased numbers of cloud droplets since low lying clouds introduce insignificant cloud greenhouse forcing, in counterdistinction with the high cirrus mechanism of Ramanathan and Collins (1991)

Discussion.
Convective transport of biogenic sulfur to the free troposphere..
The mechanism in Figure 1 uses as its "working fluid" biogenic sulfur, both particulate and gaseous, flowing through an atmospheric "loop". We provisionally consider there to be an overabundance of biogenic sulfur gases from the oceans, some fraction of which is incorporated into the "loop" as necessary working fluid to stabalize the climate. . As sea surface temperature increases, convection generalizes both in arial extent and vertical depth, carrying greater quantities of biogenic gases to the middle troposphere. Though we cannot very accurately assess the dependency of this process as a function of sea surface temperature, Clarke (1993b) and Clarke et. al, (1996) have measured enhancements in submicron aerosol in the middle tropical troposphere over the oceans. On the basis of their volatility, the aerosols seemed to be composed of droplets of sulfuric acid and ammonium sulfate and ammonium bisulfate and which appear to be distributed in a self-preserving size distribtuion (Raes, 1995).
The convective uplifting process transporting biogenic sulfur in the marine boundary layer to the high and middle troposphere, termed the "staubsauger", or "vacuum-cleaner" mechanism by Chatfield and Crutzen, (1984) is orders of magnitude more efficient than turbulent eddy diffusion in transporting boundary layer air to the free troposphere. Dickensen et al. (1987) verified in aircraft experiments that convection indeed carries substansive fluxes of trace constituents from the boundary layer to the free troposphere. They found that concentrations of tracers were enhanced in the outflow region of the storm up to 11 kilometers in altitude (Pickering et al., 1988, 1990 ; Pickering and Dickerson, 1989). Chatfield and Crutzen (1984) modeled the convective uptake of sulfur dioxide. Fluxes of sulfur species to the high troposphere increased superlinearly with convection driving energy. Unfortunately little data are available relating the fluxes of material injected into the free troposphere in response to variations in sea surface temperature. However, the variation in relative quantitiy of biogenic sulfur deposited to the free troposphere might well be rather large. Williams (1994), for example, found that the frequency of lightning in tropical deep convective regions increases rapidly with increasing sea surface temperature, the increase being evidently exponential doubling for approximately every 0.5 degrees C increase in SST.

Though the studies cited illustrate the high efficiency of deep convection in carrying boundary layer trace constituents to the free troposphere, we have found it difficult to quantify the fluxes of biogenic sulfur compounds injected to the FT as a function of sea surface temperature: there simply are so few measurements of sulfur compounds, gaseous and particulate, in the middle troposphere at clean tropical oceanic regions. Based on the long time (more than 20 years) record of aerosols at the Mauna Loa Observatory, we are able to place an upper limit of about 30 percent enhancement in sulfate aerosol during El Nino events, which are associated with widescale sea surface temperature enhancements of the order of one degree C.

Nucleation and Evolution of Sulfate Particles in the FT.
Observations of high concentrations of ultrafine sulfate aerosols in the free troposphere imply that "new" particles are nucleating from the parent biogenic gases entering by convection. Creation of new embroys by homogeneous nucleation processes is difficult to achieve in the presence of pre-existing aerosol because of the tendency for acid to condense on preexisting aerosols (e.g. Hoppel, 1975; Kreidenweis, et al., 1991; Lin, et al., 1992; Shaw, 1989). However, the process may occur in limited space-time scales to create "bursts" of new embroys. For this to occur one requires highly scavenged and humid regions such as may be found in the vicinity of heavy precipitation (Bigg et al., 1984; Covert et al., 1992; Hegg, 1991; Hegg et al., 1990, 1991; Shaw, 1989). Perry and Hobbs (1994) have detected fine particles flowing out of the detrainment region in the anvils of convective cumulus clouds and near the tops of the clouds. Clarke (1993b) documented large number of "fine" particles, of a few nanometers diameters, appearing near the Intertropical Convergence Zone (ITCZ). Brock et al. (1995) detected a persistent veil of particles with number concentrations of 103-104 cm-3 near the top of the tropical troposphere. These particles were 15 nm in diameter and appeared to be composed of sulfuric acid. In contrast, they found evidence that particles throughout the mid-latitude free troposphere are less abundant by an order-of-magnitude. The analysis of particle profiles by Brock et al. are consistent with our allegation that significant particle nucleation rates are associated with the upper regions of deep convection.

To illustrate the large numbers of particles which can be produced in brief times and in small air parcels , we lay out in Table 1 results of calculations showing the number of nucleated particles created by homogeneous, binary nucleation under different acid molecule production rates and relative humidity. It is assumed that the airmasses were well scavenged and contain no significant pre-existing aerosol surface area. Evidently significant numbers of embryos (104-109 cm-3) may be generated in hours by binary homogeneous nucleation of acid molecules and water vapor (Kreidenweis and Seinfeeld, 1988: Shaw, 1989). Embroys produced in such "bursts", undergo rapid coagulation and growth. In times of about a day they approach a self-preserving distribution like that shown in Figure 2, of a few hundred to a few thousand particles cm-3, with peak number concentration dn/dlog d, located near a diameter of 20 to 50 nm. . Such quasi steady state aerosol size distributions are quit consistent with those we measured in the free troposphere. Raes (1995) gives additional discussion along the lines proposing that the free tropospheric aerosol size distribution may be a "self preserving" ..

Residence times and chemistry of aerosols in the free troposphere.

The residence time of aerosols in the free troposphere has been estimated to be 5-15d. (Clarke, 1993a; Clarke et al., 1996). During their residence, residual DMS and SO2 travelling in the free troposphere "loop" in Figure 1 oxidizes (likely with hydroxyl radical) to convert sulfate which condenses out on the aerosols (Raes, 1992), causing them to grow. This picture of the aerosol chemistry is consistent with observed increasing concentrations of sulfur dioxide with altitude (Maroulis et al. ,1980), up to concentrations of approximately 0.1 microgram per cubic meter, and with Rodhe and Isaksen (1980) suggestion of an oxidation lifetime of 4-10 days for SO2 in the free troposphere. Thus sufficient time exists during the passage of the biogenic sulfur compounds through the loop to react large fractions of the biogenic gaseous compounds out to the surface of the aerosol.

The mass loading of the average aerosol that we measured during down slope conditions at Mauna Loa Observatory (Figure 3; Table 2) is 0.1 to 0.05 µg m-3, similar to mid-tropospheric DMS concentrations in the MBL (Andreae et al., 1988; Berresheim et al., 1990; Ferek et al., 1986).
Entry of free tropospheric aerosols into the mbl
The thermostatic loop we are suggesting necessitates that the aerosols produced by nucleation and growth processes in the free troposphere, and whichlater subside into the marine boundary layer, contribute to the nuclei for cloudy condensation in the water-rich marine boundary layer. This requires that local production of new particles within the mbl is small in comparison with particles entering by subsidence. We have found evidence that this is the case in experiments conducted in the the clean tropical Pacific.

Figure 3 and Table 2 illustrate the aerosol size distributions within and above the marine boundary layer at Hawaii. The MBL data aerosol size distribution data were taken on the east point of the island of Hawaii during times when clean air was sampled in the Trade winds. The free tropospheric data were measured from the Mauna Loa Observatory at altitude 3.3 km during clean downflow conditions at night. Free tropospheric aerosols are distributed by size monomodally , while those measured in the Marine Boundary Layer are bi-modal. This bimodal structure has been discussed by Hoppel (19xx). The larger mode consists of condensation nuclei from evaporated cloud systems within the MBL, which have grown larger by aquesous conversion of sufate from sulfur dioxide reacting with hydrogen peroxide and ozone. The "gap" between the pre-exisitng aerosol and the larger cloud-processed aerosol is at 60 nm diameter, corresponding to the activation diameter for a soluble salt, such as ammonium sulfate, at a supersaturation of approximately 1 %.

Cloud Activation Model
We discuss here results of a cloud activation model to make a preliminary inquiry about the sensitivity of cloud droplet number concentration on the CCN sulfate aerosol that subsides to the marine boundary layer. We envision a region of the marine boundary layer temporarily cleaned out of pre-existing aerosols by strong precipitation scavenging, followed by evaporation of the cloud and subsidence of free tropospheric sulfate particle distributed by size and numberconcentration according to that measured at Mauna Loa and illustrated in figure 3. We then investigate the activation of the tree tropospheric aerosol in the marine boundary layer, assuming the population of particles is cooling by adiabatic expansion in an updraft of 50 cm/sec. It is assumed that the particles are ammonium sulfate, an assumption consistent with the volatility of the free tropospheric particles that we measured previously at the Mauna Loa observatory. The cloud activation model activates approximately fifty percent of the available particles from the free troposphere (see Table 3). Particles larger than 60 nm were activated, consistent with the occurence of the "Hoppel gap" that we measured for the bimodal aerosol size distributuions in the marine boundary layer.
Activation of the cloud by this hypothetical first generation of cloud is followed by rapid aqueous in-droplet conversion of MBL sulfur dioxide to sulfate by reactions with hydrogen peroxide and tropospheric ozone (Hoppel et. al., 1994a; Hegg et al., 1990, 1991). About 90% of clouds forming, evaporate rather than precipitate (MacDonald, 1958; Pruppucher and Jaenicke, 1995). Upon evaporation the cloud condensation nuclei are returned to the atmosphere, increased in mass and diameter due to the sulfate conversion, creating a bimodal structure (Hoppel et. al, 1994a,b; Clarke et al., 1996) with a "Hoppel" gap between the "activated" and "unactivated" population. Our modeling of subsequent condensation-evaporation cycles slightly increases the width of the Hoppel Gap (by about 10 percent) and the mean diameter of the large aerosol mode, but the number concentration is preserved. Subsequent condensation cycles activate aerosols in the larger of the two modes shown in Figure 3b. This "freezes in" or "fixes" the cloud microphysics, which are responsible for determining the cloud's albedo. The bi-modally distributed particles are eventually removed by deposition to the sea, and by precipitation scavenging and are replenished by new particles entering from the free troposphere by subsidence. This assumption has been verified by measuring size and volatility of particles in clean Trade wind regime at Hawaii. Sea salt contributes only approximately ten percent of the droplet number concentration observed in marine cloud and appearence of embroys by homogeneous nucleation is apparently rare in the clean MBL: we observed only occasional ultrafine particles 5 to 10 nm diameter.

Response of cloud microphysics to increased flux of subsiding aerosols
We have run the model outlined in the last section for hypothetical increased flux of aerosol particles subsiding into the marine boundary layer from the free troposphere. This indeed affected the cloud activation process and resulted in increased numbers of cloud droplets. Larger numbers of cloud drops would slow down drizzle by reducing droplet coagulation (Twomey, 1980 ; Albrecht, 1989) and possibly modify cloud thickness and horizontal extent (Pincus and Baker, 1994), but we ignored these effects. Their inclusion would introduce even further cooling, thereby amplifying the sensitivity of the control system thermostat"
We ran the cloud activation model to estimate the response of cloud droplet concentration to increased CCN under the following two scenarios:

A)Assume that the shape of the size distribution remains constant and that the concentration in every "bin" in the aerosol number size distribution increases by the same proportion. This would be plausible if an increased sea surface surface temperature preferentially leads to a larger abundance of new embryonic particles produced by homogeneous nucleation.

B)Assume that the number of particles remains invariant, but that more gaseous-phase aerosol precursor leads to more sulfate condensation on the ultrafine particles as they descend through the troposphere. This may occur as the result of acid vapor produced by the conversion of sulfur dioxide condensing on the particles by Maxwellian diffusion, giving rise to a parabolic growth rate, dr2/dt=constant. Thus, the radius rate of change, dr/dt, during condensational growth is inversely proportional to the particle radius, r.

We know of no measurements of free tropospheric sulfate aerosol to establish the magnitude of the variations of sulfur flux travelling through the free troposphere with sea surface temperatures. We have, however, inspected the long (more than twenty year) time series of the aerosol optical scattering coefficient measured at the Mauna Loa Observatory, selecting only those times when the station was sampling aerosol during downslope conditions, and attempted to relate variations to the widescale elevated sea surface temperatures during El Nino events. There were no very apparent correlations. The additional sulfur during times of higher sea surface temperatures may be buried in the "aerosol noise", and if so cannot be larger than approximately a 30 percent increase for an approximately one degree C elevation in SST. The "aerosol noise" at the Mauna Loa Observatory is generated by crustal aerosols that periodically get swept to the central Pacific following large scale severe dust storms in Asia (Shaw, 1979 and outgassing from the fumeroles on Mauna Loa volcano. Here we adopt this "upper limit", and assume that a one degree rise in sst may increase free tropospheric CCN by as much as 30%.

The modeling of the cloud activation process, using the supersaturation spectrum calculated from the measured ensemble of free troposphere aerosol distributions at Mauna Loa Observatory, indicate that increases in Nc of 15 to 30 percent require increases in sulfur aerosol numbers by 15 to 30%, but would require larger fractional percentages of additional aerosol mass, (assume that the aerosol number concentration is conserved and aerosols simply grow larger through enhanced condensation of acid molecules).
A 30 percent increase in droplet number concentration rises cloud albedo for a globally-averaged reference cloud by 1.6% (Charlson et. al, 1987). Adopting a 30% fractional St, Sc coverage, the globally-averaged albedo would increase by a factor of 0.5% in response to a worldwide water cloud droplet number increase of 30%. This corresponds to a decrease in relative solar constant of delta alpha/(1-alpha), which, using a global average planetary albedo of 0.3 would be equivalent in reducing the sun's brightness by 0.7%. The corresponding change in global temperature may be estimated from climate models. GCM modeling suggests that the "climate sensitivity", T/(S/S), where S is the solar constant, has a value in the probable range 1.5 to 3 K/%-1. Reducing the sun's brightness by 0.7% would correspond to a temperature alteration "forcing" of 1 to 2 degrees C, or enough to "close" the loop.

Discussion and Future Research Needs
An important and interesting feature of the proposed thermoregulating mechanism is that though it employs biogenic products for its operation, we do not appeal to any environmental dependency (say on sunlight, or temperature) of the biota themselves. Biology is only incidentally involved in the loop we propsoe: the biota having co-evolved with the climate stabilizing mechanism as suggested by Schneider and Londer, 1984) in their book "Co Evolution and Life". As a working hypothesis we assume the biota are always producing sufficient sulfur precursor for the system to work.
There is an importnat need to quantify the dependence on sea surface temperatures of fluxes of sulfur species transported to the mid and high troposphere. Long-term aerosol data sets such as those measured at Mauna Loa would seem to be useful for quantifying the dependence. But the "signal" is likely to be relatively small in comparison to fluctuations from long-range transport of desert dust (and possibly industrial pollutants) from the Asian continent.

If a sulfur-aerosol thermostat is operating as we suggest, then what perturbation might arise to this system by man's industrial activities? In particular what might happen if large fluxes of anthropogenic sulfur compounds from rapidly industrializing nations like China and India get injected into deep convective systems and carried aloft to the free troposphere?. The thermostat would "conclude", because of the "logic" in which it is operating in a feedback loop, that the oceans are warming and "respond" by introducing a forced radiative cooling, but this would cool down the oceans and inhibit deep convection. This is a worrisome scenario because this may affect the strength and/or the spatial distribution of tropical monsoons. The associated systems of deep convective and heavy precipitation are prerequisite for feeding more than half of the world's population.
Acknowledgments
We thank Russell Schnell, of the Mauna Los Observatory for measurements and a reviewer for helpful suggestions. Support from NASA grant 2818-AERO92-0049 is gratefully acknowledged.

Table 1 -Particle production in the upper troposphere by homogeneous nucleation

Acid Production Rate, q (cm-3 s-1)	Maximum acid vapor density, n1 (cm-3)	Number of embryos created (cm-3)	Charactoristic Time (h)
Relative Humidity = 90%
104	2.3 x 106	6.3 x 107	1.8
105	1.3 x 108	1.1 x 108	0.3
106	7.4 x 109	2.0 x 108	0.05
Relative Humidity = 80%
104	2.2 x 10 2	4.1 x 108	11.4
105	1.2 x 104	8.7 x 108	2.4
106	6.8 x 105	1.3 x 109	0.4

Table 2 -Properties of the clean aerosol systems observed in the Free troposphere at Mauna Loa Observatory and in the Marine Boundary Layer at Kapoho, Hawaii.

Aerosol System	Total Number (cm-3)	Surface Area (µm2 cm-3)	Volume (µm3 cm-3)	‡*
Free Troposphere	313	9.79	0.27	3.47
Marine Boundary Layer	1049	42.3	1.39	3.34

* ‡ = An-1/3 V-2/3, where n, and V are the aerosol area, number, and volume concentrations respectively. This dimensionless parameter has a numerical value of XXX for a self-preserving aerosol distribution (McMurry and Friedlander, 1978). Table 3 -Cloud activation modeling results. Cloud condensation nucleus (CCN) supersaturation spectra were calculated from measured aerosol size distributions at the Mauna Loa Observatory during clean downslope conditions. The results for the "unperturbed" aerosol system (assumed to be that measured) and for 10, 30, and 1000% (factor of ten) hypothetical increases in sulfate available for the aerosol.

Aerosol System	Smax (%)	Nd (cm 3)
Scenario A - The measured number of aerosol in each bin increases
FT measured at MLO	0.303	100
FT measured at MLO + 10%	0.292	110
FT measured at MLO + 30%	0.279	121
FT measured at MLO x 10	0.130	552
Scenario B - The measured aerosol mass in each bin increases
FT measured at MLO	0.303	100
FT measured at MLO + 10%	0.298	104
FT measured at MLO + 30%	0.291	107
FT measured at MLO x 10	0.232	150

Figure Captions

Figure 1 - Schematic diagram of the proposed thermo-regulating system.

Figure 2 - Coagulation of 3 nm embryos determined by the MAEROS2 algorithm. This represents the expected change in particle size distribution as the embryos descend from 10 km to 4 km. Initial number of embryos = 1 x 106 cm-3.

Figure 3 - Aerosol number distribution measured in the marine boundary layer, Kapoho Beach, HI (upper graph). The lower graph shows data measured in the free troposphere at Mauna Loa Observatory (altitude 3.6 km).