SEARCH

SEARCH BY CITATION

Keywords:

  • general circulation model;
  • cloud condensation nuclei;
  • sulfate;
  • aerosol microphysics

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

[1] To better represent the indirect effect of aerosols on climate, a size-resolved simulation of aerosol microphysics, size distributions, number and mass concentrations has been incorporated into the GISS general circulation model (GCM). The TwO-Moment Aerosol Sectional (TOMAS) microphysics model used here conserves aerosol number as well as mass. It has high size resolution, 30 bins between 0.01 and 10 μm diameter. As a first application, a size-resolved simulation of sulfate has been performed. The model reproduces important features of the atmospheric aerosol such as number concentrations that increase with altitude and land-sea contrasts in aerosol number concentrations and size distributions. Comparisons with observations show that simulated size distributions are realistic and condensation nuclei (CN) concentrations agree with observations within about 25%. Predicted cloud condensation nuclei (CCN) concentrations are also in reasonable agreement with observations, although there are locations for which agreement would be improved by including other aerosol components such as sea salt and carbonaceous aerosols. Sensitivity scenarios show that uncertainties in nucleation and primary emissions from fossil fuels can have significant effects on predictions of CN and CCN concentrations.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

[2] It is now accepted that anthropogenic aerosols have the potential to perturb the Earth's climate by increasing cloud reflectance [Twomey, 1974; Albrecht, 1989; Pincus and Baker, 1994]. This is hypothesized to occur when anthropogenic activities increase the number of aerosol particles that serve as nuclei upon which cloud droplets form (cloud condensation nuclei or CCN). The consequent increase in cloud droplet number concentrations (CDNC) leads to brighter clouds with longer lifetimes. The resulting increase in cloud reflectance is termed indirect aerosol radiative forcing. Uncertainty in the magnitude of the indirect effect has plagued efforts to quantify the sensitivity of climate to anthropogenic perturbations such as increased concentrations of greenhouse gases. The Intergovernmental Panel on Climate Change (IPCC) has estimated that the global and annual average indirect aerosol radiative forcing is somewhere between 0 and −2.0 W m−2, as compared with +2.5 W m−2 imposed by changes in greenhouse gases [IPCC, 2001]. Moreover, this estimate neglects potential changes in cloud lifetime, including only the effect of aerosols on cloud brightness.

[3] The key process in the indirect effect is the activation of aerosol particles, typically with diameters in the range of 0.1 to 1 μm, to form cloud droplets. The cutoff diameter between CCN particles and those too small to activate depends on the degree to which the atmosphere is supersaturated with water vapor. Ideally, one would like models to predict the number of CCN for any supersaturation, which requires models that predict the aerosol size distribution.

[4] Until recently, global aerosol models did not have the capability to explicitly predict aerosol size distributions based on first principles. It is useful to think in terms of a “first generation” of global aerosol models that predict only the total mass of a given aerosol constituent without providing any description of the size distribution or number concentrations. Sulfate is the aerosol component that has received the most attention in global model studies [Erickson et al., 1995; Langner and Rodhe, 1991; Penner et al., 1994; Pham et al., 1995; Chin et al., 1996; Feichter et al., 1996; Chuang et al., 1997; Feichter et al., 1997; Kasibhatla et al., 1997; Lelieveld et al., 1997; Kjellstrom, 1998; Restad et al., 1998; Roelofs et al., 1998; Koch et al., 1999; Barth et al., 2000; Rasch et al., 2000]. Global modeling studies have also been reported for carbonaceous aerosols [Cooke and Wilson, 1996; Liousse et al., 1996; Kanakidou et al., 2000; Tegen et al., 2000], sea salt [Tegen et al., 1997; Takemura et al., 2000; Gong et al., 1998], and mineral dust [Tegen and Fung, 1994; Tegen and Lacis, 1996; Dentener et al., 1996]. With the exception of mineral dust and sea salt, these studies do not include any resolution of the aerosol size distribution.

[5] As a result of the limited information provided by first-generation aerosol models, estimates of the magnitude of the indirect effect have relied on other information to predict CCN and cloud droplet number concentrations. A common approach is exemplified by Boucher and Lohmann [1995], in which model predictions of sulfate mass were used to predict CDNC based on observed correlations. Another approach is to parameterize the size distribution of sulfate in terms of the amount of anthropogenic sulfate added to preexisting particles by condensation of gas-phase sulfuric acid and aqueous oxidation of sulfur dioxide [Chuang and Penner, 1995; Chuang et al., 1997].

[6] The empirical approach of Boucher and Lohmann [1995] has strengths and weaknesses. Probably its chief strength is that, by imposing an observed sulfate-CDNC relationship, it constrains the simulation to produce more or less realistic cloud optical properties. Another important advantage is that it requires prediction of only sulfate mass concentrations and many such models exist. The empirical approach has resulted in much-needed estimates of the magnitude of the indirect effect and associated uncertainties [Lohmann and Feichter, 1997; Kiehl et al., 2000].

[7] However, as pointed out by Kiehl et al. [2000], there is not a single, unambiguous relationship between cloud droplet number and sulfate mass that can be assumed based on observational evidence. That study used several different empirical relationships with resulting estimates of the indirect effect ranging from as low as −0.40 W m−2 to −1.78 W m−2. Most likely, the sulfate-CDNC relationship varies significantly from one part of the troposphere to another as aerosol composition and size distribution change. Besides the uncertainties inherent in the empirical approach, it has the disadvantage of concealing the physical processes that control CCN concentrations, limiting the amount of physical insight one gains from the model. Imposing a fixed sulfate-CDNC relationship means that one cannot easily test the sensitivity of model behavior to uncertainties in specific processes such as nucleation. Finally, extrapolating an observed sulfate-CDNC relationship to other locations, times of the year, or into the future, may introduce errors that are difficult to quantify.

[8] All this points to the importance of a “second generation” of global aerosol models that take a mechanistic approach to predicting CCN concentrations. In such a second-generation model, aerosol microphysical processes such as nucleation, condensation, coagulation, cloud-processing, and deposition that determine aerosol size distributions, number concentrations, and CCN concentrations are explicitly simulated. The mechanistic approach gives additional physical insight into the nature of the indirect aerosol effect, permits testing a wide variety of sensitivity scenarios, and may result in more refined estimates of indirect radiative forcing. However, adding such physical detail requires additional model evaluation to ensure number concentrations and size distributions are accurately predicted.

[9] Incorporating size-resolved aerosol microphysics into a global model is a challenging task and only recently have efforts been reported. A modal aerosol microphysical algorithm has been used to predict CCN concentrations and estimate the magnitude of the indirect effect in the MIRAGE model [Ghan et al., 2001a, 2001b]. A size-resolved aerosol microphysical algorithm has also been incorporated into the NARCM model [von Salzen et al., 2000]. This model uses 12 size sections covering particle diameters from 0.01 to 41 μm and includes sulfate and sea salt. It has been used to predict CCN concentrations over a portion of the Northern Hemisphere. A drawback is that the model uses a single-moment sectional method that makes it difficult to accurately and efficiently conserve aerosol number concentrations.

[10] This paper describes the incorporation of a two-moment sectional aerosol microphysics algorithm into a global climate model, the Goddard Institute for Space Studies General Circulation Model (GISS GCM II-prime). This provides a mechanistic simulation of global aerosol number concentrations, size distributions, and CCN concentrations. In comparison to MIRAGE, the sectional approach used here allows much greater size resolution, although this model does not yet take into account other significant aerosol types included in that work, such as organic carbon, elemental carbon, and sea salt. In comparison to the NARCM work, the model used here covers the entire globe, albeit with a coarser grid resolution. The NARCM model also includes sea salt, which is absent from this work. The two-moment sectional algorithm used here has the additional benefit of explicitly conserving aerosol number concentrations.

[11] The work presented here is limited in scope in that it considers only sulfate, neglecting important aerosol constituents such as nitrate, sea-salt, mineral dust, soot, and organic carbon. Modeling sulfate microphysics in the absence of these components is somewhat artificial because it neglects microphysical interactions between components such as sulfuric acid condensing onto sea-salt particles [O'Dowd et al., 1997]. Therefore, this paper presents only a first, but nontrivial, step in model development toward simulating aerosol microphysics and predicting CCN concentrations in global models. It will be important in the future to extend this work by including other aerosol types. For this reason, we present only a limited comparison of aerosol properties predicted by this model with observations, focusing on areas of the atmosphere where sulfate is a major aerosol component. We will also compare the results of this sulfate-only, size-resolved aerosol simulation with those from a bulk sulfate simulation in the same GCM [Koch et al., 1999] to evaluate how incorporating size-resolved microphysics and deposition influences the model's sulfur cycle.

[12] The following section describes the essential features of the size-resolved microphysical algorithm used for this work. Section 3 discusses how this algorithm has been coupled to the general circulation model. The main results of the model, CN and CCN concentrations and aerosol size distributions, are presented in section 4. A number of sensitivity scenarios have also been performed to assess model sensitivity to important uncertainties and to test how various processes control the aerosol size distribution. These are presented in section 5. Finally, section 6 discusses comparisons of model predictions with observations before conclusions are presented in section 7.

2. Formulation of Aerosol Microphysics

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

2.1. Two-Moment Sectional Algorithm

[13] To account for the aerosol microphysical processes of coagulation and condensation/evaporation in a global model, we report here the development of a TwO-Moment Aerosol Sectional (TOMAS) microphysics model. The algorithms used in the TOMAS model are adaptations of cloud microphysics algorithms [Tzivion et al., 1987, 1989] to aerosol processes. Although many sectional, or bin, algorithms exist for treating size-resolved aerosol microphysics, the unique advantage of those we use here is that they track two independent moments of the aerosol size distribution for each size bin or category. The two moments usually selected, which we also adopt in this work, are aerosol number and mass defined as follows,

  • equation image

and

  • equation image

where Nk and Mk are the total number and mass of aerosol particles in the k-th size category, nk(x) is the number of particles with masses between x and x + dx, and xk is the lower boundary of the k-th size category. Nk and Mk are independent prognostic variables in the sense that the ratio of mass to number of a given size bin, or average particle mass in that bin, is allowed to vary in space and time.

[14] It is important here to stress some of the differences between this approach and more common single-moment algorithms. Typically, single-moment algorithms choose aerosol mass as the prognostic variable and generally do not conserve aerosol number. When modeling the indirect effect, however, number concentrations are important and one requires an algorithm that conserves aerosol number. This has led to the development of single-moment algorithms that conserve number by shifting aerosol mass between size categories as appropriate [Russell and Seinfeld, 1998]. This approach is hampered by the discretization of the aerosol size spectrum implicit in single-moment schemes; the ratio of mass to number in a given size category is fixed to a constant value at all times such that the average particle mass cannot change. As a result, conserving aerosol number in a single-moment scheme is achieved only by sacrificing realism in the aerosol size distribution.

[15] For example, imagine a small amount of gaseous sulfuric acid condensing onto a large number of monodisperse particles such that the growth of any given particle is small compared to the difference between size categories. Realistically, one would expect all the particles to stay within the original size bin, which would gain mass. The unchanged number concentration, however, implies an increase in average particle size for that bin. A single-moment, number-conserving algorithm in which average particle size for each category is fixed cannot accommodate this solution, however. Instead, aerosol number is conserved, in effect, by placing all the condensing mass onto a few particles that hop to the average size of the next size category. While total number concentrations are conserved, the aerosol size distribution is unrealistically distorted. This kind of numerical diffusion is a problem if one wants to predict CCN concentrations as these are a function not just of total aerosol number, but also of the size distribution.

[16] In comparison, a two-moment sectional algorithm has a number of advantages. Physically, such an algorithm has the desirable property of predicting and conserving aerosol number, the important quantity for estimating the indirect aerosol effect. Another advantage is that it is more accurate than single-moment algorithms with the same size resolution [Tzivion et al., 1987] or faster than single-moment algorithms with the higher resolution required to attain similar accuracy. Finally, the additional degree of freedom in a two-moment algorithm make it possible to simultaneously conserve aerosol number and minimize numerical diffusion, whereas single-moment algorithms require a tradeoff between these two goals.

[17] Our application of the two-moment algorithm to the GISS GCM II-prime uses a moving sectional approach in which the boundaries between size bins are defined in terms of dry aerosol mass. The moving sectional approach is advantageous because condensation and evaporation of aerosol water in response to changes in atmospheric relative humidity do not move aerosol between size categories, a process that otherwise results in unwanted numerical diffusion. We use 30 size bins with the lower boundary of the smallest category at 10−21 kg dry aerosol mass per particle, close to 0.01 μm dry diameter for a typical aerosol density of 1.8 g cm−3. Each successive boundary has double the mass of the previous such that the upper boundary of the largest category is about 10 μm diameter, or 10 size bins per decade of aerosol diameter.

[18] This is rather high size resolution compared with other global or continental-scale models of aerosol microphysics that either use modal algorithms [Ghan et al., 2001a, 2001b] or fewer size bins [von Salzen et al., 2000]. The additional size resolution should be advantageous in the future when coupling the aerosol microphysical simulation to the GCM cloud scheme and predicting the number of activating particles for different supersaturations. Even with the high size resolution, the model runs reasonably fast because of the computational efficiency of the two-moment algorithm, an aerosol microphysics solver that uses adaptive time steps, and the somewhat coarse spatial resolution of the GISS GCM II-prime. The model requires 1 h on a single processor of an SGI Origin 2100 to simulate one day of model time, or two weeks to simulate one year, and further optimization may be possible.

2.2. Aerosol Physical Properties

[19] At various times in the model, it is necessary to compute physical properties of the aerosol such as particle size and density. For simplicity, we assume that all sulfate exists uniformly as ammonium bisulfate, a global average chemical composition [Adams et al., 1999]. A water uptake curve for ammonium bisulfate at 273 K has been calculated offline from the GCM, varying the relative humidity from 1 to 99% using the aerosol thermodynamic equilibrium module, ISORROPIA [Nenes et al., 1998]. We use this curve to specify the amount of aerosol water, whenever necessary, for current conditions of relative humidity in a GCM grid cell. The density of the resulting ammonium bisulfate-water mixture is calculated using the measurements of Tang and Munkelwitz [1994].

2.3. Coagulation

[20] Coagulation of particles in the atmosphere is an important sink of aerosol number and a mechanism by which freshly nucleated particles grow to larger sizes. It is accounted for using the method developed by Tzivion et al. [1987]. Details of the derivation are given in that work, which we summarize briefly here. The stochastic collection equation (the portion of the aerosol general dynamic equation relating to coagulation) is recast using the definitions of the two moments given in Equations (1) and (2), resulting in a set of equations that describe how aerosol number and mass in each size category evolve with time. Assuming that the variation of the coagulation coefficient over a single size section is negligible, the resulting equations for the moments are,

  • equation image

and

  • equation image

where ξ depends on the bin spacing and equals 1.0625 for the mass-doubling used here, Kj,k is the coagulation coefficient for particles in the j-th bin with particles in the k-th, fk and ψk are parameters that describe the linear approximation to the number distribution that are defined in Tzivion et al. [1987], xk is the lower boundary of the k-th size bin in terms of dry mass, mk is the average mass of particles in the k-th bin, and I is the total number of size bins.

[21] To calculate the coagulation kernel, Kj,k, we assume that aerosol particles coagulate via Brownian diffusion, neglecting the effects of gravitational settling and turbulence. The kernel is recalculated at every grid cell and time step to take into account changes in particle size. We calculate the kernel based on the average hydrated particle sizes in bins j and k, neglecting the variation of the kernel with particle size within a given bin. Diffusivities of aerosol particles are calculated from the Stokes-Einstein formula. For particles smaller than about 0.1 μm, it is necessary to correct for noncontinuum effects, which we do using the expression given by Dahneke [1983]. Equations (3) and (4) are solved using an adaptive time step where the time step is limited such that the aerosol number or mass concentration in any size category with numerically significant aerosol loading does not decrease by more than 25% or increase by more than an order of magnitude.

2.4. Condensation/Evaporation

[22] Condensation of gas-phase sulfuric acid to existing aerosol particles is an important source of sulfate aerosol mass and a means by which small particles grow to CCN size. We use the algorithm presented by Tzivion et al. [1989] to simulate this process. This algorithm uses an analytical solution for the rate of growth of individual particles to compute yk, the initial mass of a particle before condensation that grows to a mass of xk. Therefore, a particle whose starting mass falls between yk and yk + 1 will fall into the k-th size bin after condensation. The mass and number moments at the end of the condensational time step are then computed from the following conservation equations,

  • equation image

and

  • equation image

where τ is a parameter that describes the driving force for condensation,

  • equation image

In the above equation, D is the diffusivity of sulfuric acid in air, M is its molecular weight, R is the ideal gas constant, ρk is the density of aerosol in size category k, and F is a correction factor described below. The quantity Δp is the difference between the partial pressure of sulfuric acid and its equilibrium vapor pressure. For sulfuric acid, the equilibrium vapor pressure is assumed to be negligible.

[23] The correction factor, F, in Equation (7) takes into account noncontinuum effects for condensation to small particles as well as the possibility that not all gas-phase sulfuric acid molecules that encounter the surface of a particle stick to that particle. We use the expression proposed by Dahneke [1983] in which F is a function of Kn, the Knudsen number and α, the accommodation coefficient. We have chosen a value of 0.65 for the accommodation coefficient of sulfuric acid [Pöschl et al., 1998].

[24] To improve the efficiency of the condensation algorithm, Equation (7) is written to take into account the decrease in the partial pressure of sulfuric acid as it condenses onto aerosol. This approach allows longer condensational time steps and faster solution. An analytical solution that describes how Δp changes with time is derived in Tzivion et al. [1989] and used in Equation (7).

[25] As with the coagulation algorithm, we use an adaptive time step to efficiently solve the equations for condensation. The length of the time step is chosen such that individual particles in any size bin do not grow by more than 10%, the partial pressure of sulfuric acid does not fall below 25% of its original value, and is never longer than 15 min.

3. Coupling Aerosol Microphysics to the GCM

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

[26] The previous section presented a general description of the TOMAS aerosol microphysics model itself; here we present details regarding its coupling to the GISS GCM II-prime. This description relates to a single model experiment designated, somewhat arbitrarily, as the “base case”. A number of sensitivity scenarios have also been realized with the model, and their unique features will be described later.

3.1. GISS General Circulation Model

[27] The TOMAS model described above is run online in the GISS GCM II-prime [Hansen et al., 1983; Rind and Lerner, 1996]. The version used here has a horizontal resolution of 4 degrees latitude and 5 degrees longitude and nine vertical layers between the surface and the model top at 10 mb. Typically, most of these layers are in the troposphere with only one or two in the stratosphere. The time step for tracer processes is 1 h. Advection of heat, moisture, and chemical tracers are all calculated using a quadratic upstream scheme. Sea surface temperatures are specified based on climatological averages.

[28] The GCM uses two distinct parameterizations to simulate stratiform and convective clouds. Details regarding stratiform clouds are given in Del Genio et al. [1996]. This treatment carries liquid water content as a prognostic variable and assumes that stratiform clouds appear in a GCM grid cell once the average relative humidity exceeds 60%. The parameterization of sub-grid scale convective clouds predicts the convective mass flux necessary to neutralize the buoyancy instability at cloud base [Del Genio and Yao, 1993]. It explicitly includes two updrafts, one entraining and one nonentraining. One third of this upward flux is balanced by a downdraft and the rest by large-scale subsidence. After the one-hour model time step, liquid water in convective clouds either precipitates, evaporates, or detrains from the convective plume to form a persistent stratiform cirrus anvil.

3.2. Sulfur Cycle Model

[29] This study builds upon an earlier sulfate model, without size resolution, described in Koch et al. [1999], hereafter referred to as K99. Gas-phase tracers included in this version of the model are H2O2, DMS, SO2, and H2SO4. In addition, 30 aerosol tracers each for both aerosol sulfate mass and aerosol number concentration are required to represent the moments of each size section in the two-moment microphysics algorithm. Finally, a single bulk aerosol tracer representing MSA is included as it was in K99. Therefore, the model tracks a total of 65 tracers.

[30] The model was integrated for a total of 25 months of simulation time for the base case and all sensitivity scenarios. All tracer concentrations were initially zero, and the first 13 months were discarded to allow the model to relax from these initial conditions. Unless otherwise noted, all results shown here are annual averages over the remaining 12 months.

[31] A difference between the sulfur emissions used in K99 and those used in the base case scenario here is that we assume that all anthropogenic sulfur emissions from the GEIA inventory are in the form of sulfur dioxide while K99 assumed that 3% were in the form of sulfate. The effects of particulate emissions are evaluated in a separate sensitivity scenario (see section 5.2). Otherwise, emissions are treated as in K99.

[32] The model of sulfur chemistry described in K99 is maintained here, except that a gas-phase sulfuric acid tracer is introduced. Whereas in K99, the gas-phase sulfuric acid produced by the reaction of sulfur dioxide with the hydroxyl radical was assumed to condense immediately to form aerosol sulfate, here we distinguish between particulate and gaseous S(VI). This is necessary to simulate the competition between condensation of sulfuric acid onto existing aerosol particles and nucleation of new particles. Although we distinguish between H2SO4(g) and sulfate aerosol, the rate of the gas-phase reaction between sulfur dioxide and the hydroxyl radical is the same as in K99. The size distribution of sulfate produced in cloud is described below.

3.3. Nucleation

[33] As nucleation is a negligibly small pathway for gas-to-particle conversion in terms of aerosol mass, it was entirely neglected in the first generation of global aerosol models that predicted only bulk aerosol mass. Nucleation, however, is important in determining aerosol number concentrations and size distributions. Therefore, in an aerosol microphysical model such as this, it is essential to realistically model this process.

[34] However, a number of problems complicate any treatment of nucleation in a global aerosol model. First, there are significant uncertainties surrounding both the mechanisms and rates of new particle formation in the atmosphere. While substantial efforts have aimed at understanding binary nucleation in the H2SO4-H2O system [Doyle, 1961; Mirabel and Katz, 1974; Heist and Reiss, 1974; Jaecker-Voirol and Mirabel, 1989; Kulmala and Laaksonen, 1990; Laaksonen et al., 1995], theoretical estimates [Coffman and Hegg, 1995; Korhonen et al., 1999] suggest that, when ammonia also participates in a ternary system, nucleation rates can be orders of magnitude higher. Furthermore, atmospheric measurements [Clarke et al., 1998b; Kulmala et al., 2000] in which nucleation is observed at sulfuric acid concentrations below those required by a binary nucleation mechanism suggest that ternary nucleation or other mechanisms are at work in the atmosphere. Other studies have suggested that biogenic and anthropogenic organic species [Marti et al., 1997] or ion-ion recombination [Turco et al., 1998] may play a role in new particle formation.

[35] Second, even given a particular mechanism of new particle formation and an accurate rate expression for that mechanism, the rate is highly sensitive to variations in temperature, relative humidity, gas-phase precursor concentration, and aerosol surface area. This presents serious challenges for global models with their coarse spatial resolution that cannot resolve potentially important sub-grid variability in these quantities. For example, nucleation has been observed to occur in regions of cloud outflow where scavenging has reduced aerosol surface area and enhanced actinic flux near clouds creates conditions where sufficiently high concentrations of gas-phase sulfuric acid are allowed to accumulate [Clarke et al., 1998a, 1999b]. A recent modeling study [Liu et al., 2001] similarly predicted new particle formation in regions of cloud outflow and also found that the simulated nucleation rate depended strongly on the horizontal and temporal resolution of the model. As another example, studies have shown that fluctuations in temperature and vapor pressures resulting from atmospheric waves and turbulence have the potential to enhance nucleation rates by orders of magnitude over those that would be predicted based on large-scale average conditions [Easter and Peters, 1994; Nilsson and Kulmala, 1998; Nilsson et al., 2000]. These examples suggest that directly applying expressions for nucleation rates, accurate for a given point in space, to entire GCM grid cells using large-scale average quantities may well be inaccurate. This will be the case if the nucleation rate is a nonlinear function of parameters with substantial subgrid variability, as is generally true.

[36] Given the uncertainties and difficulties just discussed, we choose to incorporate a simple treatment of new particle formation in the GISS GCM II-prime and to determine the sensitivity of model behavior in a separate scenario. In this treatment, we allow gas-phase sulfuric acid one model time step (1 h) to condense onto existing aerosol particles. At the end of the time step, the remaining gas-phase sulfuric acid concentration is compared against a critical concentration necessary for significant new particle formation,

  • equation image

where Ccrit is the critical sulfuric acid concentration in μg m−3, T is temperature in K, and RH is relative humidity on a scale of 0 to 1. Equation (8) is a fit [Wexler et al., 1994] to nucleation rate calculations [Jaecker-Voirol and Mirabel, 1988] using a new particle formation rate of 1 cm−3 s−1 as the criterion for rapid nucleation. If the sulfuric acid concentration at the end of the time step exceeds the critical value, the excess mass is assumed to nucleate. The mass and number of the smallest size bin are increased such that the sulfuric acid concentration is reduced to Ccrit. Otherwise, no nucleation occurs. Although such a description of new particle formation in the atmosphere is obviously far from perfect, this stage of model development does not warrant a more sophisticated treatment. We will examine the sensitivity of model behavior to such uncertainties in a sensitivity scenario with enhanced nucleation.

3.4. In-Cloud Sulfur Oxidation

[37] Production of sulfate by the reaction in cloud droplets of sulfur dioxide with hydrogen peroxide is important both because it is the largest source of atmospheric sulfate and because it strongly influences the aerosol size distribution. This occurs as particles activate into cloud droplets, gain solute via in-cloud chemistry, and often evaporate, resulting in particles that are larger than before cloud processing. In this version of the model, we retain the treatment of in-cloud chemistry described in K99, which assumes that the rate of formation of sulfate can be described by a bulk chemistry formulation. Therefore, any effects on the rate of sulfate formation stemming from the differing chemical compositions of distinct cloud droplets are neglected in this work. Also, oxidation of sulfur dioxide by ozone is not included in this model.

[38] In the model, sulfate produced by aqueous oxidation of sulfur dioxide is distributed across only those size sections representing particles large enough to activate. We assume that the GCM's stratiform clouds experience a maximum supersaturation of 0.19% such that particles with dry diameters larger than 0.082 μm activate. Similarly, particles larger than 0.033 μm activate in the more vigorous convective clouds, corresponding to a supersaturation of 0.75%. In order to partition the newly produced sulfate across the various activated size sections, we make two assumptions. First, we assume that particles from any given activated size section grow to the same size on average as those from any other size section. Second, we assume that the rate of sulfate production in a given size section is proportional to the volume of cloud water associated with that section. Therefore, we distribute the newly formed sulfate across the activated size sections in proportion to the number of particles in that size section.

[39] To account for the growth of some particles from their initial size bin to a larger size category by cloud processing, we use the same algorithm as described above for condensation/evaporation. In this case, we choose the τ parameter defined above in Equation (7) for each size bin such that the sulfate added to that size bin, before accounting for transfer of particles to larger bins, is proportional to the number of particles initially in that category. That is, we calculate τj for the j-th size section such that,

  • equation image

where

  • equation image

where mij is the average mass of a single particle in size section j before cloud processing, mfj is the average mass of a particle in size section j after cloud processing, Δm is the total amount of new sulfate produced in that particular GCM grid cell during the current model time step, and Nact is the total number of activated particles in that grid cell.

[40] Although certain aspects of the in-cloud chemistry treatment presented here are crude, in particular the assumptions of a single supersaturation for a given cloud type and bulk cloud chemistry, we feel that this is all that is warranted at the current time. Future work will refine these assumptions as more sophisticated treatments of aerosol activation and cloud microphysics and chemistry are developed for the GISS GCM II-prime.

3.5. Size-Resolved Dry Deposition

[41] Dry deposition of gaseous species and the bulk aerosol MSA tracer is calculated as in K99. For size-resolved sulfate mass and number, however, a size-dependent parameterization of dry deposition has been implemented. The scheme uses a version of the resistance in series approach [Wesely and Hicks, 1977] developed by Chin et al. [1996] and modified as in K99. For aerosols, the resistance in series approach has to be modified slightly to account for gravitational settling of particles [Seinfeld and Pandis, 1998, Equation 19.7]. The surface resistance is assumed to be zero for aerosols. The aerodynamic resistance is calculated from GCM surface momentum and heat fluxes as in K99 and has no size dependence. The resistance in the quasi-laminar layer is calculated from Equation 19.18 of Seinfeld and Pandis [1998].

[42] Figure 1 shows annual-average dry deposition velocities that result from this method as a function of size bin. These are shown with the dry diameter of the size bin on the horizontal axis although the computed deposition velocities in fact depend on the amount of water uptake by the particles at any instant. Curves are shown for global average values as well as from a forested region with high deposition velocities. The deposition velocities exhibit a minimum in the 0.1 to 1 μm diameter size range. As most of the aerosol mass resides in this size range, these low velocities determine the removal of sulfate mass by dry deposition. Note that these tend to be lower than those used in K99. Consequently, dry deposition is much less important as a sink of sulfate mass in this model than in previous studies. For comparison, Gallagher et al. [1997] review and report measurements of dry deposition velocities from forested regions. Typical values are 0.1 cm s−1 for 0.1 μm particles increasing to about 1 cm s−1 for 1.0 μm particles, approximately an order of magnitude higher than the forested values shown in Figure 1. On the other hand, measurements in wind tunnels report lower deposition velocities, consistent with those in Figure 1 [Sehmel, 1980]. At this point, reasons for the discrepancies remain unclear although Gallagher et al. [1997] discuss some hypotheses.

image

Figure 1. Global and annual average dry deposition velocities (cm s−1) as a function of particle size (solid curve). Same, but for a forested region (45°E-70°E, 52°N-68°N, dashed curve). For comparison, the global and annual average dry deposition velocity used for bulk sulfate in K99 is also shown (straight, dashed line).

Download figure to PowerPoint

3.6. Size-Resolved Wet Deposition

[43] Precipitation removes aerosol from the atmosphere when particles activate to form cloud drops that later precipitate and when falling raindrops impact and scavenge particles below cloud. Removal of aerosol from the atmosphere via wet deposition is treated here in much the same way as in K99. A complete description is given in that work; here it suffices to say that they assume that all aerosol mass in convective plumes is fully dissolved in cloud water via nucleation (in-cloud) scavenging and use a first-order rate loss parameterization to describe this process in stratiform clouds. Subsequently, dissolved aerosol generally mimics the GCM's cloud liquid water in its behavior with a fraction that is removed via wet deposition as precipitation reaches the ground, another fraction evaporating (potentially in a different GCM vertical layer), and some detraining from convective clouds. A first-order removal is also applied to aerosol below precipitating clouds to simulate below-cloud scavenging. Here we have made simple changes to this treatment to account for the size dependence in these processes.

[44] For nucleation scavenging, we assume that only those particles large enough to activate are dissolved in cloud liquid water, using the same activation criteria described earlier in the section on aqueous oxidation. Smaller particles are not removed by wet deposition in the base case scenario, but can be carried with convective plumes, downdrafts or subsidence to different vertical layers in the same way as undissolved gases. The fraction of aerosol that activates and is subject to wet removal accounts for essentially all the aerosol mass, such that sulfate mass concentrations and lifetimes with respect to wet deposition are not significantly different than in K99. The main effect of this assumption is to exempt a significant fraction of total aerosol number from wet removal. This treatment of in-cloud scavenging neglects the possibility that inactivated aerosol particles in clouds, so-called interstitial aerosol, may be incorporated in cloud water and removed by wet deposition if they collide with cloud drops. Therefore, we may underestimate removal of smaller, Aitken mode particles in the base case scenario. This possibility is examined in a sensitivity scenario in which interstitial aerosol is scavenged.

[45] Below precipitating clouds, we use the same first-order removal rate formulation discussed in K99. In that work, aerosol mass was scavenged below cloud with a 0.1 mm−1 washout rate constant. Here we use a size-dependent washout rate constant taken from Figure 2 of Dana and Hales [1976]. Their Figure 2 shows theoretical washout rate coefficients as a function of aerosol size for both monodisperse and polydisperse aerosol size distributions. As our size bins are fairly narrow, we apply the appropriate rate constant taken from the monodisperse curve to each size bin.

4. Simulated Microphysics of Tropospheric Sulfate

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

4.1. Sulfate Mass

[46] Table 1 shows the sulfur budgets calculated with various scenarios of the microphysical model and compares them to the one calculated with K99's model that lacked size-resolution. Except for the bulk dry deposition scenario, each will be discussed in detail in section 5, so we only briefly summarize them here. The primary emissions scenario assumes that 3% of the GEIA sulfur emissions occur as particulate sulfate with a large number of particles emitted in the 10 to 30 nm diameter size range. These particles represent direct emissions of sulfate, new particle formation in subgrid scale plumes, or could be considered surrogates for other types of primary emissions such as organic carbon. In the enhanced nucleation scenario, the critical concentration of sulfuric acid required for new particle formation is lowered by an order of magnitude compared to the base case. Bulk dry deposition applies a uniform dry deposition velocity to particles of any size. In this scenario, the deposition velocity is calculated as in K99 to facilitate comparison with that version of the model. Finally, the preindustrial scenario excludes anthropogenic sulfur emissions. The base case, primary emissions, and enhanced nucleation scenarios are collectively referred to as “modern-day scenarios” and serve to illustrate the microphysical impact of primary particles and nucleation.

Table 1. Comparison of Sulfur Budgets Computed With Size-Resolved Aerosol Microphysics and Previous Model Version Without Microphysics [Koch et al., 1999], Labeled K99
 Base CasePrimary EmissionsEnhanced NucleationBulk Dry DepositionPreindustrialK99
DMS
   Sources (Tg S yr−1)
Emissions10.810.810.810.810.810.7
Sinks (Tg S yr−1)      
Oxidation10.710.710.710.710.710.7
Burden (Tg S)0.050.050.050.050.050.06
Lifetime (d)1.71.71.71.71.71.9
 
SO2
   Sources (Tg S yr−1)
Emissions72.870.872.872.83.870.8
DMS oxidation9.79.79.79.79.79.6
Total82.580.582.582.513.580.4
Sinks (Tg S yr−1)      
SO2 + OH15.314.815.315.32.713.1
SO2 + H2O227.927.527.927.97.531.6
Dry deposition37.936.737.937.93.235.5
Wet deposition1.51.41.51.50.10.2
Total82.680.582.682.613.580.4
Burden (Tg S)0.680.660.680.680.080.56
Lifetime (d)3.03.03.03.02.32.6
 
SO42−
   Sources (Tg S yr−1)
Emissions0.02.00.00.00.01.9
Nucleation0.050.040.160.050.030.00
Condensation of H2SO415.214.815.115.22.613.1
SO2 + H2O227.927.527.927.97.531.6
Total43.244.343.243.210.146.6
Sinks (Tg S yr−1)      
Dry deposition0.81.00.97.80.29.2
Wet deposition42.343.342.335.39.937.4
Total43.244.343.243.210.146.6
Burden (Tg S)0.770.800.780.630.150.73
Lifetime (d)6.56.66.65.45.55.7
 
MSA
   Sources(Tg S yr−1)
DMS oxidation1.01.01.01.01.01.1
Sinks (Tg S yr−1)      
Dry deposition0.20.20.20.20.20.2
Wet deposition0.90.90.90.90.90.9
Total1.01.01.01.01.01.1
Burden (Tg S)0.020.020.020.020.020.02
Lifetime (d)7.57.57.57.57.57.6

[47] All of the modern-day budgets, with or without size-resolved aerosol microphysics, are similar. Because the K99 results were compared extensively with observations, this gives us confidence that our simulation of sulfate mass is realistic. The agreement between the various size-resolved simulations in terms of sulfate mass disguise, however, important differences in predicted CN and CCN concentrations that will be discussed in greater detail below. One minor difference between these results and those from K99 is the somewhat higher wet deposition of sulfur dioxide reported here. This is because K99 counts sulfur dioxide oxidized to sulfate and immediately removed via wet deposition in the same cloud cycle as sulfate deposition whereas it is counted as sulfur dioxide deposition in this work.

[48] A more significant difference is dry deposition of sulfate, with the rates reported here being an order of magnitude lower than those in K99. This can be explained by recalling Figure 1, which shows that dry deposition velocities are smallest in the accumulation mode size range where most sulfate mass resides. For reference, the dashed horizontal line shows the average dry deposition velocity used in K99. Running the aerosol microphysical model with the K99 dry deposition parameterization applied to all size sections results in a sulfate budget that agrees with K99 as shown in the “bulk dry deposition” scenario. The slower dry deposition in this work accounts for the higher lifetime and burden of sulfate as well as the increased wet deposition of sulfate, which partly compensates for the slower dry deposition.

[49] A small caveat is that a more realistic size-resolved simulation, one that includes coarse particles such as dust and sea salt and allows sulfate to partition onto those particles, would result in higher average deposition velocities. To the extent that sulfate in the atmosphere partitions onto coarse particles, the current treatment will underestimate dry deposition.

4.2. Aerosol Number Concentrations

4.2.1. Vertical profiles

[50] Figure 2 shows the vertical profiles of sulfate mass, CN number, and CCN (0.2%) number concentrations at STP conditions (273 K and 1 atm). Results are shown for the three modern-day and one preindustrial scenarios. Of the several scenarios realized, these four were chosen for demonstrating important factors influencing aerosol behavior in the global troposphere. Each point in the vertical profile represents the global and annual average for that model layer. The decreasing sulfate concentration with altitude is familiar and is easily explained by the dominance of sulfur dioxide emissions at the surface and aqueous oxidation to produce sulfate in boundary layer clouds. Of the various modern-day scenarios performed as part of this study, we show only the primary emissions scenario as the others differ by less than 10% in terms of sulfate mass.

image

Figure 2. Annual average vertical profiles of (a) sulfate mass (μg m−3), (b) CN concentrations (cm−3 at STP), and (c) CCN (0.2%) concentrations (cm−3 at STP) for four model scenarios: preindustrial (lightweight dashed line), base case (lightweight solid line), enhanced nucleation (heavyweight dashed line), and primary emissions (heavyweight solid line).

Download figure to PowerPoint

[51] In contrast to sulfate mass, CN concentrations generally increase with altitude. This vertical profile results from the higher rates of nucleation in the upper troposphere and tropopause region and is consistent with observations (see section 6.4). An important caveat is that the magnitude of this trend depends on the details of the nucleation mechanism. It might have been less strong, for example, had we implemented a ternary nucleation mechanism, as the presence of ammonia would have tended to enhance new particle formation primarily in the boundary layer. An important exception to this overall trend is the local maximum in particle concentrations at the surface in the primary emissions scenario. This maximum is also consistent with observations so the primary emissions scenario must be considered the most realistic in this respect.

[52] Number concentrations of CCN, on the other hand, exhibit a decrease with altitude, similar to sulfate mass. The correlation with sulfate mass is not surprising because particles have to be relatively large in order to function as CCN and also because in-cloud production of sulfate, the largest source of sulfate, contributes further to the mass of CCN particles. However, CCN number concentrations differ from sulfate mass concentrations in that they exhibit significant differences from one modern-day scenario to another. These sensitivities to model formulation will be discussed in greater detail below. For now, it is enough to mention that, although CCN number and sulfate mass are correlated, the correlation is not absolute, and a given sulfate mass may be consistent with significantly different CCN concentrations. The exact nature of the sulfate-CCN relationship depends, of course, on the details of the aerosol size distribution and varies not just between different model scenarios, but also with space and time in the Earth's atmosphere. An aerosol simulation with explicit microphysics, such as the one presented here, allows one to capture and explore this kind of variability in ways that would be difficult or impossible in a model with a fixed, empirical sulfate-CCN relationship.

4.2.2. Zonal averages

[53] Figure 3, in which zonal average CN and CCN number concentrations (cm−3 at STP conditions of 273 K and 1 atm) are presented, demonstrates some other features of the atmospheric aerosol. For example, strong convection in equatorial regions has significant impacts on aerosol microphysical behavior. In all scenarios, CCN concentrations are significantly depleted in the tropical upper troposphere as convective precipitation removes most of the available CCN. A direct consequence is the large numbers of CN in roughly the same region, but shifted slightly to higher altitudes. The cold temperatures of the upper troposphere and low aerosol surface areas and high humidity associated with regions of cloud outflow create ideal conditions for new particle formation. This behavior is consistent with the results of models with more detailed cloud dynamics and microphysics [Zhang et al., 1998] and also with observational evidence from recent field campaigns in the eastern Pacific [Clarke et al., 1999a, 1999b].

image

Figure 3. Zonal and annual average CN and CCN (0.2%) concentrations (cm−3) for four model scenarios. Contours of CN concentrations are 0, 10, 20, 50, 100, 200, 500, 1000, 2000, 5000, and 10,000. Contours of CCN (0.2%) concentrations are 0, 1, 2, 5, 10, 20, 50, 100, 200, 500, and 1000.

Download figure to PowerPoint

[54] Contrasts between hemispheres are also evident in Figure 3. These are not very pronounced for CN number concentrations with the important exception of the primary emissions scenario. This suggests that new particle formation is somewhat insensitive to gas-phase anthropogenic emissions, but primary emissions have impacts on CN concentrations. On the other hand, strong contrasts between hemispheres in CCN concentrations are evident in all modern-day scenarios. In the boundary layer, average Northern Hemisphere CCN concentrations are 275 cm−3 compared with 65 cm−3 in the Southern Hemisphere. One can conclude that the impact of anthropogenic sulfur dioxide emissions on CCN concentrations results less from forming new particles than contributing extra sulfate mass to a pool of existing particles such that they grow to sizes where they can activate in clouds.

4.2.3. Latitude-longitude maps

[55] Figure 4 presents predicted annual average CN and CCN concentrations (cm−3 at STP conditions of 273 K and 1 atm) in the lowest model layer. One feature of the atmospheric aerosol that is captured is the higher number concentrations over land than over oceans. This can be seen in all the boundary layer results, but is most pronounced when primary emissions are included. Predicted CN concentrations exceed 10,000 cm−3 in the most polluted industrialized regions of the Northern Hemisphere as a consequence of primary emissions. In other scenarios, continental CN concentrations typically fall in the 200–1000 cm−3 range. In the most remote marine areas, predicted CN concentrations in the base case are lower, generally 100–200 cm−3. Marine concentrations exceed 200 cm−3 in other situations, however, such as areas closer to land or in the enhanced nucleation and primary emissions scenarios.

image

Figure 4. Latitude-longitude map of annual-average (a) CN and (b) CCN (0.2%) concentrations (cm−3) in the model surface layer for four model scenarios. Values in the upper-right corners represent layer averages.

Download figure to PowerPoint

[56] Comparing the four scenarios presented in Figure 4 suggests some conclusions. First, the contrast between the primary emissions scenario and the others indicates that primary particles make a significant, even dominant, contribution to aerosol number concentrations in the polluted continental boundary layer. A comparison of the base case and preindustrial scenarios, on the other hand, shows very little difference. In other words, gas-phase emissions of sulfur dioxide increase particle concentrations somewhat in the upper troposphere, where the nucleation process is active, but have little impact on CN concentrations in the boundary layer.

[57] CCN concentrations also tend to be higher over land than over the oceans, a contrast that the model also captures well. Figure 4 shows that, over land, CCN (0.2%) concentrations in the primary emission scenario are greater than 100 cm−3 nearly everywhere and exceed 1000 cm−3 in industrialized areas. Leaving aside the extremes of highly polluted and very remote, annual average CCN (0.2%) concentrations in most continental areas fall in the range of 200–1000 cm−3. The primary emissions scenario, therefore, gives CCN concentrations in good agreement with expected values. The two other modern-day scenarios, on the other hand, predict lower CCN concentrations, from 100–500 cm−3. These are somewhat lower than expected, so the primary emissions scenario seems to be the most realistic in this regard. Similarly, this scenario gives marine CCN (0.2%) concentrations mostly in the 10–200 cm−3 range, also in good agreement with expected values. The other modern-day scenarios also have a similarly realistic range of marine CCN concentrations, although they exceed 100 cm−3 much less frequently.

[58] Primary emissions, therefore, dramatically impact both CCN and CN concentrations in polluted regions of the boundary layer. Comparing scenarios with and without them clearly indicates that including a source magnitude on the order of 3% of anthropogenic sulfur emissions is necessary to adequately represent continental CN and CCN concentrations. Their impact on marine areas is also noticeable throughout much of the Northern Hemisphere.

[59] In comparison to the modern-day scenarios, the preindustrial scenario has very low CCN concentrations, typically around 30 cm−3 and rarely exceeding 50 cm−3. Important natural aerosols that contribute to CCN concentrations such as sea salt and biogenic organics are not accounted for, however, so this cannot be considered to be a realistic simulation of the preindustrial atmosphere. Still, the magnitude of the difference in CCN concentrations between preindustrial and modern-day scenarios suggests that anthropogenic emissions of sulfur have significantly changed CCN concentrations.

4.3. Factors Controlling CN and CCN Concentrations

[60] Figure 5 shows sources and sinks of aerosol number for three modern-day and one preindustrial scenario. This plot demonstrates which processes control aerosol number concentrations and lifetimes and serves as a rough estimate of the uncertainties associated with estimated sources and sinks. The values presented in this figure were derived by taking global and annual average source and sink rates and converting to volumetric rates assuming an average tropopause height of 12 km.

image

Figure 5. Global and annual average sources and sinks of aerosol particles for four model scenarios. Values in boxes are average CN concentration (cm−3) and particle lifetime (days).

Download figure to PowerPoint

[61] The upper part of the figure gives a rough idea of the range of estimated rates for the production or emission of new particles to the atmosphere. The base case scenario, using binary nucleation theory, results in an average production rate of about 6 × 10−4 cm−3 s−1. The corresponding rate for the sensitivity scenario with enhanced nucleation is higher by a factor of three. Other plausible parameterizations of nucleation might yield a still larger range of rates. Note that the specified rate implicitly depends on the lower limit of the aerosol size spectrum, 10 nm diameter in this case.

[62] The primary emissions scenario indicates that primary particles are a source of new aerosol number comparable to, or perhaps substantially larger than, the nucleation source. However, substantial uncertainties exist. Our scenario assumed that 3% of anthropogenic sulfur emissions are particulate sulfate, but others have estimated this fraction at 5% [Langner and Rodhe, 1991]. It appears that this fraction has only been estimated with enough precision to verify that it makes a small contribution to sulfate mass concentrations. One or two percent by mass of anthropogenic sulfur emissions, however, represents a significant uncertainty in the number of emitted particles.

[63] These estimates would, of course, be different in a more comprehensive model that included other aerosol types. Including primary emissions of black carbon, sea salt, and mineral dust would tend to suppress nucleation in the atmosphere, but the emissions of these species are also less well quantified than sulfate. The budget of aerosol number is, therefore, largely unconstrained, and future experimental, field, and modeling studies should seek to better understand and quantify these processes.

[64] Sinks of aerosol number show a similar range of magnitudes across the various scenarios, as they must if they are to balance the source terms. Wet and dry deposition of particles are relatively unimportant sinks in terms of aerosol number. By far the dominant sink globally for particle number is coagulation. Because its rate depends on the square of aerosol number, coagulation acts as a powerful control against large fluctuations in number concentrations. Consequently, even the threefold increase in nucleation between the base case and the enhanced nucleation scenario results in only a twofold increase in average CN concentrations. Another example of how coagulation influences aerosol number concentrations can be seen by comparing the enhanced nucleation scenario with the primary emissions scenario. Although they have the same source magnitude to within 25%, the resulting average CN concentrations differ by a bit more than 65%. In the primary emissions scenario, the particle source is concentrated in the polluted boundary layer, and coagulation efficiently reduces number concentrations resulting in the low particle lifetime. The enhanced nucleation source, in contrast, is spread more diffusely throughout the upper troposphere such that coagulation is less efficient and the particle lifetime longer.

[65] In the base case scenario, the average CN concentration is approximately 500 cm−3, with a lifetime of around 10 days. For the modern-day scenarios with higher source terms, higher concentrations result. The nonlinearity inherent in the process of coagulation explains the shorter lifetimes in these scenarios. In comparison, it is interesting to note that the lifetime of sulfate mass is 6.5 days for each of the three modern-day scenarios shown in Figure 5. The lifetime of individual aerosol particles, however, varies from 5 to 10 days in the various scenarios. That is, the lifetime of aerosol number can be higher or lower than that of aerosol mass, depending on whether condensation or coagulation dominates.

[66] The base case scenario may be taken as an example of a situation in which condensation and cloud processing dominate and the lifetime of aerosol number is higher than for aerosol mass. An idealized particle trajectory in this scenario would go as follows (see Figure 6). A particle is formed in the upper troposphere during a nucleation event and grows by condensation of sulfuric acid. Particle concentrations are relatively low and coagulation is unimportant, such that the particle floats around the free troposphere, maintaining a small size of 10–40 nm diameter. Eventually, after several days, the particle grows to sufficiently large size to activate and is entrained into the boundary layer. It now grows principally by in-cloud oxidation of sulfur dioxide to a much larger size, 0.2-0.4 μm, before being removed from the atmosphere by wet deposition after 6 days or so. In this condensation-dominated case, the particle had a lifetime of around 10 days while most of the sulfate mass had a shorter lifetime of around 6 days. This trajectory is similar to ones suggested by others in which new particle formation in the free and upper troposphere and subsequent subsidence are an important source of aerosol number to the boundary layer [Dinger et al., 1970; Hoppel et al., 1973; Bigg et al., 1984; Clarke, 1993; Raes et al., 1995]. Alternatively, the primary emissions scenario exemplifies a situation in which coagulation determines the particle lifetime. In this case, a primary particle is emitted in a polluted region with high number concentrations such that it coagulates with another after a couple of days. The mass associated with that particle, however, persists in the atmosphere for several more days.

image

Figure 6. Idealized particle trajectories illustrating the relative lifetimes of aerosol number and mass under different microphysical scenarios. The upper section shows a trajectory in which condensation and cloud processing dominate, similar to the base case scenario. The lower section shows a trajectory in which coagulation dominates, similar to the primary emissions scenario.

Download figure to PowerPoint

[67] The preceding discussion of the aerosol number budget applies to CN number concentrations and is, in fact, dominated by the behavior of particles 10 to 30 nm in size. One is also interested, of course, in the factors that control CCN number concentrations. The dominant source of particles large enough to serve as CCN is condensation of gas-phase sulfuric acid onto smaller particles until they grow to sizes that can activate in clouds. This mechanism is effective in the 30 to 100 nm diameter size range. In-cloud oxidation, on the other hand, is the dominant mechanism by which particles grow larger than 100 nm in size. In strong contrast to the results presented above for CN number, the main sink for particles larger than 30 nm is wet deposition. Between 30 and 100 nm, both wet deposition and coagulation are of roughly equal importance as sinks of particle number. Above 100 nm, coagulation is not a significant source or sink of particles.

[68] In summary, the picture of aerosol microphysics that emerges from this simulation is already familiar. Ultrafine particles are produced in the atmosphere by nucleation. In terms of number, most are removed by coagulation, but some survive and grow by gas-phase condensation to CCN size. Subsequently, cloud processing is the dominant growth mechanism. Finally, CCN are removed mostly by wet deposition.

4.4. Size Distributions

[69] In this section we present and describe simulated number size distributions for different regions of the atmosphere, focusing on results from the primary emissions scenario as these are considered to be the most realistic.

4.4.1. Marine size distribution

[70] Figure 7a shows the annual average aerosol number distribution in the remote South Pacific (180° W to 120° W and 60° S to 0° S). The model reproduces the bimodal nature of the marine aerosol size distribution, broadly consistent with the range of observations reviewed by Fitzgerald [1991]. As explained by Hoppel et al. [1985, 1986], the two modes correspond to a smaller, inactivated “nuclei” mode and a larger accumulation mode of CCN particles. The gap between the modes results from cloud processing. An important feature of the marine aerosol that is missing here is a coarse mode of sea salt particles. Future work will focus on adding sea salt and other types of aerosol to the model. The total number of particles predicted by the model is 190 cm−3. Moreover, the data of Fitzgerald [1991] show that, in the Southern Hemisphere, there are fewer particles in the larger mode while the two modes are of roughly equal size in the Northern Hemisphere. Although not shown here, examining the model's predicted size distributions in corresponding marine areas shows that it reproduces this trend.

image

Figure 7. Model predictions of annual average aerosol number size distributions taken from the primary emissions scenario for (a) a remote marine location, (b) polluted boundary layer locations, and (c) free and upper troposphere.

Download figure to PowerPoint

4.4.2. Polluted continental size distribution

[71] Figure 7b shows size distributions from three polluted continental regions: North America (100° W to 70° W and 28° N to 48° N), Europe (0° E to 25° E and 40° N to 56° N), and East Asia (105° E to 145° E and 24° N to 48° N). Once again, the model produces a bimodal aerosol size distribution with modes at 15 and 150 nm. In the model, particle concentrations in polluted areas range from an average of 6600 cm−3 in East Asia up to 15,000 cm−3 in Europe, with North America intermediate at 9000 cm−3. Qualitatively, the model correctly places the majority of the particles in the smaller mode. Primary emissions play a dominant role here, and the model is sensitive to the crude assumptions made here regarding the number and size distributions of primary particles.

4.4.3. Free and upper tropospheric size distributions

[72] Figure 7c shows aerosol size distributions above the boundary layer that represent global averages over individual model layers. The model predicts, in agreement with observational evidence shown later, a strong vertical concentration gradient in the free troposphere, so we show results at 634, 468, 321, and 201 mb. In contrast to the marine boundary layer, these size distributions are unimodal, not having experienced significant cloud processing. Global and annual average free tropospheric number concentrations are predicted to be 350–375 cm−3 (ambient conditions) throughout the free troposphere. The size distribution at 201 mb is heavily skewed toward small particles in the 10 to 20 nm size range. This is the area of the atmosphere in which nucleation is most active, at least as parameterized in this model, and this process has a dominant effect on the size distribution. Sulfur dioxide concentrations are low in the upper troposphere, so these freshly nucleated particles have little opportunity to grow to larger sizes. Moreover, larger particles are not likely to be carried aloft by deep convection as they serve as the nuclei for cloud droplets and tend to be removed by precipitation.

5. Sensitivity Scenarios

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

5.1. Enhanced Nucleation

[73] Earlier, we discussed the difficulties inherent in trying to represent new particle formation in a global aerosol microphysical model, acknowledging the imperfection of the nucleation parameterization used here. Here we determine the sensitivity of the microphysical simulation to an increase in nucleation by reducing the critical concentration of sulfuric acid required for nucleation from Equation (8) by a factor of ten. This approach is motivated by observations in which nucleation is observed at sulfuric acid concentrations an order of magnitude below those predicted by binary nucleation theory [Clarke et al., 1998b]. Although not shown here, a separate sensitivity scenario with an accommodation coefficient of 0.1 yielded similar results.

[74] First, it is worth noting that the change in the nucleation parameterization has a minor effect on predicted sulfate concentrations. The impact is highest at the tropopause, where sulfate concentrations are low and nucleation rates high, but is still less than 10%. The changes in the budget of aerosol number are significant, however, and illustrate the robustness of atmospheric aerosol number concentrations with respect to nucleation. For example, an order of magnitude change in the critical concentration required for nucleation results in a three-fold increase in average nucleation rates. This, in turn, results in CN concentrations that are a factor of two higher, and CCN (0.2%) concentrations generally increase by only 10–35% depending on which part of the atmosphere one examines. There are at least two reasons for this chain of diminishing impacts from nucleation to CCN concentrations. One is the, already noted, nature of the coagulation process that, being second order in number concentrations, tends to dampen any attempt to disturb them. Another is that it takes a substantial amount of sulfate mass for a particle to grow to CCN size, such that the fixed emissions of sulfur serve to limit the number of nucleated particles that can grow this large.

[75] Given the uncertainties associated with atmospheric nucleation, it is comforting that predicted CCN concentrations are relatively insensitive to the exact details of this process. One should resist, however, the temptation to ignore such uncertainties entirely. In this sensitivity scenario, CCN (0.2%) concentrations increase approximately 35% on average in the boundary layer, an increase that is certainly large enough to have a significant impact on accompanying simulations of cloud microphysics and albedo. In the free troposphere, CCN (0.2%) concentrations increase by 10–20%.

[76] Given the noticeable impacts the enhancement in nucleation rates has on model behavior, especially with respect to CN concentrations in the upper troposphere, a natural question is whether such changes are realistic or not. In a later section, we will show that observations of CN in the upper troposphere tend to agree better with the results from the base case scenario than with those from this sensitivity scenario.

5.2. Primary Emissions From Fossil Fuel Combustion

[77] The first generation of global sulfate mass simulations did not emphasize including a realistic treatment of primary sulfate aerosol emissions. Previous global model studies of sulfate have estimated the fraction of anthropogenic sulfur emitted in the form of sulfate at 3–5% or neglected it entirely [Langner and Rodhe, 1991; Pham et al., 1995; Chin et al., 1996; Koch et al., 1999]. Greater sophistication or precision was unnecessary for calculating sulfate mass loadings as the anthropogenic contribution to sulfate mass is overwhelmingly through emissions of gas-phase sulfur dioxide and subsequent oxidation in the atmosphere to form aerosol sulfate. Although primary emissions of sulfate do not contribute significantly to sulfate mass loadings, this is not necessarily true of aerosol number. If primary emissions are a significant source of new atmospheric particles, it follows that second-generation aerosol microphysical models that predict number concentrations need to take better account of such emissions. Here we describe results of a sensitivity scenario in which primary emissions from fossil fuel combustion are included to determine their contribution to aerosol number concentrations.

[78] The subgrid scale plume immediately downwind of a power plant stack has unique chemistry on shorter length scales than can be resolved with the coarse horizontal resolution of climate models. Therefore, we use emissions estimates that represent the air some tens of kilometers downwind of a stack that begins to mix with the larger environment rather than estimates of emissions at the exit of the stack itself. Note that the only primary sulfate emissions included in this scenario are those from land-based anthropogenic combustion of fossil fuels. Primary emissions from biomass burning and aircraft are not considered in this scenario, although primary emissions from aircraft will be examined in another. As in K99, we assume here that 3% of anthropogenic sulfur emissions from the GEIA inventory [Benkovitz et al., 1996] occur as aerosol sulfate. Similar to the approach used in regional models [Binkowski and Shankar, 1995] and based on the work of Whitby [1978], we divide the primary emissions into a nuclei and an accumulation mode. Fifteen percent of the emissions by mass are assumed to occur in the nuclei mode and the balance in the accumulation mode. Emissions rates into individual size sections are based on the lognormal parameters reported by Whitby [1978]. These are geometric mean diameters of 10 and 70 nm and standard deviations of 1.6 and 2.0, respectively.

[79] As expected, primary sulfate emissions have a negligible impact on the sulfur mass budget. Compared with the base case scenario, the annual average sulfur dioxide burden decreases and the sulfate burden increases by about 3%. More interesting is the effect that primary emissions have on the budget of aerosol number. Given the mass of sulfur emitted as primary sulfate and the assumed size distribution of those emissions, the global source of primary sulfate particles from anthropogenic fossil fuel emissions is approximately twice as large as the nucleation source in the base case scenario. This confirms that primary emissions are a significant source of new particles to the atmosphere.

[80] It is interesting to note that, even with this additional source of particles, the global and annual average nucleation rate in this scenario is only 20% lower than that estimated for the base case. Nucleation continues to occur in the free and upper troposphere more or less at the same rate despite the large source of primary particles, which are mostly confined to the boundary layer. Because primary emissions are mostly confined to the boundary layer, both dry and wet deposition increase substantially, but loss of particle number via coagulation is still the dominant sink.

[81] Overall, the global average increase in CN concentrations resulting from primary emissions is only 20%, although this global average hides the enormous increase in the most polluted regions of the atmosphere (see Figure 2 and Figure 4). Primary emissions result in CN concentrations in excess of 5000 cm−3 in industrialized regions whereas they were only 200–500 cm−3 without them. Figure 2 also shows some impact in the lower portion of the free troposphere, but this region is not as sensitive to primary emissions as it is to the nucleation source.

[82] More surprising is the sensitivity of CCN concentrations to primary emissions. Compared with the base case, CCN (0.2%) concentrations are approximately double in the boundary layer and 30–50% higher in the free troposphere (see Figure 2). This is not so much a result of the few particles that are emitted already large enough to act as CCN as the ability of the increased numbers of CN emitted in polluted conditions to grow to the CCN size range. Although the fixed mass of sulfur emissions, in principle, limits the number of CN that can grow to CCN sizes, there is sufficient excess gas-phase sulfur that this turns out to not be limiting. The average particle size of the CCN mode is, in fact, smaller in the primary emissions scenario than in the base case, but still large enough to function as CCN.

[83] Although it is not useful to do a detailed comparison to observations in the polluted boundary layer with an aerosol model that includes only sulfate, a quick check of predicted CN concentrations against observations is enough to indicate that the results from the primary emissions scenario are more realistic (see section 6.4). This sensitivity scenario has shown that primary particles, besides making the dominant contribution to CN number in the polluted boundary layer, can significantly enhance CCN concentrations. This result suggests that better emissions inventories of primary particles are needed. First-generation sulfate models can adequately predict sulfate mass concentrations with only crude estimates of sulfate emissions, but the sensitivities of CN and CCN concentrations demonstrated here indicate that these are no longer sufficient. Although this sensitivity scenario considers only primary sulfate particles from anthropogenic fossil fuel combustion, one can assume that the same lessons apply to primary sulfate emissions from biomass burning and other sources as well as to nonsulfate primary particles.

5.3. Primary Emissions From Aircraft

[84] Although the number of particles emitted by aircraft is much smaller than the number emitted by fossil fuel combustion, in this scenario we test the hypothesis that their location in the upper troposphere gives them a disproportionate impact. Primary particles emitted by aircraft are, in fact, predominantly composed of black carbon, not sulfate [Brasseur et al., 1998; Petzold and Schröder, 1998; Petzold et al., 1999]. It seems likely, however, that some of the sulfur dioxide emitted by the aircraft is oxidized within the subgrid scale plume such that these particles are coated with sulfate by the time they mix with the surrounding air [Brasseur et al., 1998; Brown et al., 1996; Schumann et al., 1996], although the precise amount of the sulfur dioxide that gets oxidized is uncertain [Fahey et al., 1995; Petzold and Schröder, 1998]. In this study, we are more concerned with the contribution of these particles to tropospheric number concentrations than their chemical composition so, for simplicity, we treat them as being composed entirely of sulfate.

[85] For an estimate of total sulfur emitted as well as the spatial distribution of those emissions, we use the same inventory for aircraft as was used in K99 [Baughcum et al., 1993] and that was also used in the base case scenario. In the base case, all aircraft sulfur was emitted as sulfur dioxide. In this sensitivity scenario, we assume that 50% of aircraft sulfur is emitted as sulfate. This is on the high end of estimates for the conversion of fuel sulfur to sulfate within the plume of the aircraft, but was chosen because it represents a more realistic estimate of the number of particles, principally soot, emitted by aircraft. Even so, this amounts to approximately 3 × 1015 particles per kilogram of fuel burned, on the low end of the estimated range. Particle emissions from jet aircraft could be an order of magnitude higher. We assume a lognormal size distribution of the primary particles with a geometric mean diameter of 45 nm and geometric standard deviation of 1.5, values that are representative of the size distribution of soot emitted by aircraft [Petzold and Schröder, 1998].

[86] Changing a fraction of aircraft sulfur dioxide emissions to particulate sulfate has a negligible effect on simulated sulfate mass concentrations, a result also obtained by Kjellstrom et al. [1999]. In terms of the global budget of aerosol particles, primary emissions from aircraft are also small. For these emissions estimates, primary particles amount to less than 1% of the nucleation source, not to mention a source of primary particles from the surface of similar magnitude. Higher, but still realistic, emissions estimates would give a source that is less than 10% of the nucleation source. Changes in CN and CCN (0.2%) concentrations are also small, less than 1% in layer averages even for the free troposphere and upper troposphere. In fact, CN concentrations actually decrease slightly when primary emissions from aircraft are included. The reason for this is that, for the size distribution of emissions assumed here, there are enough larger particles to suppress nucleation slightly. Recent work on the impact of aircraft on upper tropospheric CN concentrations estimated that primary particles from aircraft increase concentrations by 6% [Anderson et al., 1999]. Anderson et al. [1999] used a much higher emissions index (5 × 1016 particles emitted per kg aircraft fuel) and did not take into account potential microphysical feedbacks, such as the suppression of nucleation by larger particles. Including this feedback suggests that primary emission from aircraft have only a very small impact on the global scale.

5.4. In-Cloud Scavenging of Interstitial Aerosol

[87] In the base case scenario, we made the simple assumption that interstitial aerosol particles, those not large enough to activate in clouds, were not subject to in-cloud scavenging. In reality, some fraction of the interstitial aerosol will collide and coalesce with cloud drops and be removed by wet deposition. Without resorting to detailed simulation of microphysical processes in clouds, a simple scenario bounds the sensitivity of the model to assumptions about the scavenging of interstitial aerosol particles. In this scenario, we assume that all such particles are fully incorporated into cloud drops and subject to the same wet deposition processes as the activated particles. The true behavior of the atmosphere lies somewhere between the two extreme assumptions of no scavenging (base case) and complete scavenging (this scenario).

[88] This sensitivity scenario predicts concentrations of sulfate mass that are nearly identical to those predicted in the base case scenario, a result of the fact that interstitial aerosol particles, being small in size, do not account for much of the total sulfate mass. CN number concentrations decrease by 10–20% on average in the boundary layer, but increase by a few percent in the free troposphere such that the global average is nearly unchanged. This is because scavenging the interstitial aerosol leads to an increase in new particle formation, mostly in the upper troposphere. Wet deposition of CN particles increases by about 50% such that particle lifetime is decreased from 10.3 days to 9.3 days. Overall, the impact on CN concentrations is small because increased rates of new particle formation largely compensate for faster removal rates. Concentrations of CCN (0.2%) decrease by about 10% throughout most of the troposphere. In summary, the two extreme assumptions made regarding in-cloud scavenging of interstitial aerosol result in CN and CCN concentrations that differ by 10% or less in most areas. Therefore, it is not necessary to incorporate detailed simulations of this process in global models.

5.5. Preindustrial Atmosphere

[89] The final model scenario represents a first approximation of aerosol microphysics in the preindustrial atmosphere, in order to make a preliminary assessment of the impact of anthropogenic sulfur emissions on CCN concentrations. We use the same assumptions regarding sulfur emissions to the preindustrial atmosphere as in K99 and Adams et al. [2001]: no anthropogenic sulfur emissions, DMS and volcanic sulfur dioxide emissions that are the same as the modern-day scenarios, and biomass burning sulfur dioxide emissions that are 10% of their modern-day values. An important caveat is that DMS emissions, the most important aerosol precursor for this scenario, are uncertain by a factor of two [Kettle and Andreae, 2000], and the estimate we use of 10.8 Tg S yr−1 is lower than others. This, and the absence of important natural aerosol types such as sea salt, biogenic organics, and dust, will tend to underestimate preindustrial aerosol concentrations and, consequently, overestimate anthropogenic impacts.

[90] Preindustrial sulfate concentrations resulting from this set of emissions assumptions have been presented before in K99. We will not discuss them except to note that the estimated modern-day sulfate burden, 0.77 Tg S, is five times the preindustrial, 0.15 Tg S. The vertical profile of average preindustrial sulfate concentrations is shown in Figure 2a. The comparison between preindustrial and modern-day scenarios depends on whether one compares against the modern-day scenario with or without primary sulfate emissions. We have already discussed the impact of primary emissions above, so here we focus on a comparison with the base case scenario without primary emissions in order to isolate the impact of anthropogenic emissions of gas-phase sulfur dioxide. Figure 5 shows that CN concentrations are 20% higher in the modern-day scenario than in the preindustrial, a result of a 65% increase in nucleation. These results suggest that anthropogenic sulfur dioxide emissions do, in fact, lead to new particle formation. Increased coagulation results in a shorter lifetime in the modern-day scenario, 10 days as opposed to 14 days, compensating somewhat for the increased nucleation. In fact, the increases in CN concentrations are most evident in the upper troposphere where they are as high as 25%. How much impact this would have on boundary layer aerosols and clouds is questionable, however, because CN number concentrations are only 5% higher by the time they work their way down to this altitude. Looking at CCN (0.2%) concentrations, on the other hand, shows more clear evidence for the potential of sulfur dioxide emissions to contribute to CCN. CCN (0.2%) concentrations are approximately two and a half times higher in the base case scenario than in the preindustrial. So, although sulfur dioxide emissions increase CN concentrations, they are more effective at increasing CCN concentrations simply by contributing extra mass to existing particles such that they grow to CCN size.

6. Comparison With Observations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

[91] Assessing the performance of a detailed model of aerosol microphysics in a GCM is challenging. Because the GCM produces its own meteorology, it does not reproduce the weather conditions of any particular time period when observations have been made. The ideal comparison would be against a long-term, climatological data set that averages over the various meteorological regimes that influence aerosol behavior. Unfortunately, such data are rare for areas in which sulfate is the major aerosol component, such as the marine boundary layer. Although detailed observations of aerosol composition and size distributions do exist, they tend to be from field campaigns lasting several weeks, not a sufficiently long period to be considered climatological. On the other hand, long-term, multiyear observations are often filter pack measurements that give no information about size distributions and number concentrations.

6.1. Marine Boundary Layer

[92] Over the last decade, ships have extensively sampled aerosol number concentrations and other properties over large regions of the Pacific Ocean as part of the MAGE92, RITS93, RITS94, and ACE1 field campaigns [Quinn et al., 1995; Covert et al., 1996; Bates et al., 1998]. Here we focus on CN12 (particles larger than 12 nm diameter) concentrations measured with a condensation particle counter during portions of the cruises over the central Pacific Ocean (155° W to 130° W and 32° S to 20° N). Taken together, these data represent approximately six months of total sampling time covering all months of the year except June through September over a period of four calendar years from 1992 to 1995. To remove some of the variability evident on short temporal and spatial scales, the data were binned into four-degree latitudinal bands corresponding to the GCM's latitudinal resolution.

[93] Average CN12 concentrations for the latitudinal bins mostly fall into a range between 200 and 400 cm−3 with an average of 325 cm−3. The data do not show any clear latitudinal gradient. The annual and regional average CN12 concentration from the base case scenario is 190 cm−3. Looking within the region, some grid cells predict annual average CN12 concentrations below 150 cm−3, while most were in a range between 150 and 250 cm−3. The base case scenario, therefore, produces CN12 concentrations that are mostly within the range of observations but are too low on average. Even assuming an additional 30 cm−3 of sea salt particles, such as would occur at a reasonably high wind speed of 12 m s−1 [O' Dowd et al., 1997], would be insufficient to close the gap. On average, CN12 concentrations from the primary emissions scenario were 215 cm−3. Primary particles, therefore, have a small, but not entirely negligible, effect on CN12 concentrations even in this remote marine region. They too, however, are insufficient to explain the tendency of the model to underpredict marine number concentrations. The enhanced nucleation scenario, on the other hand, shows much higher marine CN12 concentrations with an average of 420 cm−3, now an over-prediction of the number of particles. This occurs despite the fact that most of the nucleation in this scenario, as in the base case scenario, occurs in the upper troposphere, not the marine boundary layer.

[94] In summary, the model reasonably represents CN concentrations in the marine boundary layer. Including primary emissions from industrialized areas and assuming 30 cm−3 of sea salt particles, predicted CN12 concentrations are in the range of observed values, but tend to be too low by about 25%. A sensitivity scenario with enhanced nucleation indicates that the discrepancy may result from a slight underprediction in the rate of nucleation, although the nucleation rate in that scenario is almost certainly too high. Alternatively, the GCM may be underestimating subsidence of air with higher particle concentrations from the free troposphere, or perhaps it is necessary to invoke a ternary nucleation mechanism to account for CN concentrations in marine areas.

[95] Measurements of aerosol size distributions in the North Atlantic are available from a cruise undertaken as part of the Joint Global Ocean Flux Study (JGOFS) [Van Dingenen et al., 1995]. The cruise track followed a course from Halifax, Nova Scotia to the Canary Islands and back during September and October of 1992. Van Dingenen et al., [1995] report number size distributions for clean, modified, and polluted conditions classified using air mass back trajectories. Because this study does not account for substantial concentrations of carbonaceous aerosols that were observed under modified and polluted conditions, we compare our results against their clean size distribution. The predicted size distribution represents a September/October average over a portion of the North Atlantic (50° W to 25° W and 30° N to 45° N) where the cruise ship encountered the most clean conditions.

[96] Figure 8a shows the lognormal fit to the cruise data and the size distributions predicted in the base case, primary emissions, and enhanced nucleation scenarios. As discussed earlier, the simulated and observed marine aerosol size distributions show a bimodal character with a mode of larger, cloud-processed CCN particles and a smaller “Aitken” mode. There is a close correspondence between the sizes of the simulated and observed modes and the “Hoppel” gap between them. The small dip at 0.046 μm diameter in the simulated Aitken mode results from processing by the GCM's convective clouds, which were assumed to have a higher supersaturation than the stratiform clouds that accounts for the more prominent gap at 0.09 μm.

image

Figure 8. Model predictions of aerosol size distributions compared with observed size distributions for (a) the North Atlantic, (b) the free troposphere, and (c) tropopause. For part (a), model predictions from three modern-day scenarios are shown. In parts (b) and (c), only predictions from the primary emissions scenario, judged to be the most realistic overall, are shown.

Download figure to PowerPoint

[97] As was the case in the Pacific, the base case scenario underpredicts number concentrations, having a total of 150 cm−3 compared against the 400 cm−3 that were measured. In contrast to the Pacific, primary particles are important and the model predicts total number concentrations of 330 cm−3 when they are included. The enhanced nucleation scenario also results in particle concentrations of 330 cm−3. Once again, a combination of primary emissions and a modest enhancement in nucleation rates would result in the best agreement.

[98] The underprediction is more severe if one looks at CCN particles above 0.09 μm diameter, obviously a concern if one is interested in estimating the indirect effect of aerosols. The observations show particle concentrations of 200 cm−3 between 0.09 and 1 μm diameter. Predictions are 65, 110, and 80 cm−3 in the base case, primary emissions, and enhanced nucleation scenarios, respectively. A comparison shows that the simulated sulfate mass concentrations are similar to those observed on the cruise so it is unlikely that this is the source of the discrepancy. A more likely explanation is the absence of sea salt and other aerosol types. Although sea salt is often associated with aerosol in the supermicron size range, significant amounts are also found in the size range below 1 μm [O'Dowd et al., 1997; Murphy et al., 1998], so some sea salt particles were probably sampled and contributed to the observed CCN mode. Another important possibility is that organic aerosol is contributing a significant number of particles in this size range. Even the central region of the North Atlantic cannot be considered remote and is substantially impacted by anthropogenic activity. Evidence for this comes from observations from the cruise ship, in which measured MSA to sulfate ratios indicated that most of the sulfate was in fact anthropogenic.

[99] To summarize, the agreement between the observed and simulated shapes of the size distribution indicates that the model is successfully capturing the microphysics of marine aerosol. Sensitivity scenarios suggest that it is necessary to accurately account for nucleation in remote marine areas and primary emissions in those closer to land in order to predict accurately number concentrations. In addition, it appears that, throughout the North Atlantic, anthropogenic aerosol contributes significantly to the number of particles in the accumulation mode.

6.2. Free Troposphere

[100] Schröder et al. [2002] measured aerosol number concentrations and size distributions as part of the Lindenberg Aerosol Characterization Experiment (LACE98) in July and August of 1998. These comprise data taken during 7 flights, or about 11 hours of sampling time, in the boundary layer, free troposphere, and upper troposphere. The region sampled by the aircraft (2° E to 14° E and 47° N to 53° N) includes portions of western and central Europe, centered mostly over Germany.

[101] Figure 8b shows a comparison of predicted average number size distributions for two vertical layers in the free troposphere with the median distribution measured during the LACE98 campaign. The observations are averaged over a wide range of altitudes, from 4 to 10 km, with a total number concentration of 360 cm−3 of particles larger than 5 nm. Total number concentrations predicted by the model are 430 cm−3 and 320 cm−3 for model vertical layers centered at 468 mb (roughly, 5 to 7 km) and 321 mb (7 to 10 km), respectively. Observed CN concentrations are therefore reproduced to within 20% in this situation.

[102] Looking at the size distributions themselves, the observations show a unimodal structure that is typical of the free troposphere. There are high concentrations of Aitken mode particles with a peak around 40 nm diameter, which are products of new particle formation that is occurring in the free troposphere and upper troposphere. Freshly nucleated particles grow to these sizes by coagulation and condensation of gas-phase sulfuric acid. The lack of an accumulation mode is evidence that these particles have not undergone cloud processing to a significant degree. Predicted size distributions are similarly dominated by a peak in the Aitken mode, whose location nicely matches the observations. The model tends to produce somewhat more particles above 0.1 μm diameter. K99 noted that this model may overpredict sulfate concentrations in the free troposphere, which might explain the shift in the predicted size distributions to larger sizes. Overall, we conclude that the model accurately represents aerosol microphysical processes in the free troposphere, considering the early stage of model development as well as the relatively short sampling time of the data with which we are comparing.

6.3. Upper Troposphere

[103] In the upper troposphere and tropopause region, we compare the model predictions against two data sets. The first data set is a result of a flight from Darwin, Australia to Tokyo, Japan [Clarke, 1992]. Looking at averages over 100–200 km flight segments, measured CN15 concentrations varied between 1000 and 2000 cm−3 (STP conditions). No clear latitudinal gradient is discernible. Average CN15 concentrations predicted by the model for the same region and time period are 1200 cm−3 for the primary emissions scenario, within the range of observations. Predicted CN concentrations from the enhanced nucleation scenario, on the other hand, are 2000 cm−3 on average, toward the high end of the observations. It is also interesting to note that the observed ultrafine concentrations, defined as the difference between measured CN3 and CN15 concentrations, show clear evidence that the equatorial upper troposphere is a region with high rates of new particle formation. The current microphysical algorithm does not include particles smaller than 10 nm, so no quantitative comparison is possible, but this observation is in qualitative agreement with the predicted maximum in nucleation rates in this region.

[104] The Schröder et al. [2002] data, described in the previous section on the free troposphere, offer another set of CN concentrations and size distributions for testing the performance of the model. Figure 8c shows a comparison between the median of observations made in the 10–12 km altitude range and average predictions for the same region, altitude, and time period. The observations show median particle concentrations of 410 cm−3 (ambient conditions). Our scenario with primary emissions is in reasonable agreement at 290 cm−3, although about 25% lower than the observations. In this case, the scenario without primary emissions actually gives better agreement with average CN concentrations of 380 cm−3, although the primary emissions scenario is still considered more realistic and gives better agreement with observations overall. Similar to the result for the Clarke [1992] data, the scenario with enhanced nucleation overpredicts CN concentrations. The overprediction is, however, much more severe in this comparison with average CN concentrations of 950 cm−3.

[105] Looking at the shape of the size distributions, there is remarkably good agreement between observations and model results. The most obvious difference between the size distributions is the peak in the predicted size distribution at 15 nm. The fact that the model has a lower size boundary of 10 nm diameter and that this is a region where nucleation is producing large numbers of ultrafine particles in the 3 to 10 nm size range may account for the discrepancy. The lower size boundary may force an artificial lumping of these ultrafine particles with particles in the first and second size bins resulting in an apparent overprediction by the model for these bins. Alternatively, this discrepancy may be a statistical artifact. We are comparing average model results with the median of the observations. If nucleation events produce large numbers of small particles infrequently, the predicted average value will be larger than the observed median.

[106] The model, therefore, performs quite well in predicting CN concentrations and size distributions in the upper troposphere. It also appears that the simple parameterization of binary nucleation theory accurately predicts average new particle formation rates. Although some increase in predicted nucleation rates would correct the model's tendency to underpredict CN concentrations, very much enhancement of nucleation rates leads to an overprediction.

6.4. Vertical Profiles

[107] The CN concentrations and size distributions measured by Schröder et al. [2002] in the free and upper troposphere have already been compared to predictions. The same data set also provides CN concentrations, but not size distributions, in the polluted boundary layer. Other aerosol types besides sulfate need to be included before a detailed comparison is meaningful for the polluted boundary layer. However, a brief discussion of boundary layer CN concentrations completes the vertical profile and leads to two relevant conclusions. Measured CN14 concentrations in the polluted boundary layer exhibit a strong vertical gradient with concentrations in excess of 10,000 cm−3 at the surface and declining to 1,000 cm−3 at 2 km altitude. Above the boundary layer, CN14 concentrations reach a minimum and then increase until a local maximum is reached at the tropopause.

[108] When emissions of primary particles are included, the boundary layer gradient is reproduced more or less accurately. For this scenario, CN14 concentrations predicted for the surface layer, which is approximately 0.7 km thick, are 10,000 cm−3 and decrease to 4,700 cm−3 for the second layer that extends to 1.4 km. Direct comparison is difficult given the coarse vertical resolution of the model, but results are approximately correct with the estimates of primary particles used here. In contrast, without primary emissions CN14 concentrations are grossly underpredicted, being only 50–200 cm−3 in this same area. One can conclude that the crude assumptions made here about primary emissions are approximately correct and necessary for accurately representing aerosol microphysics in the polluted boundary layer. Further refinements that include better inventories of primary particles, incorporation of other aerosol types into the model, and more detailed comparison with observations are necessary, however.

[109] Another important feature of the observational evidence that is reproduced by the model is the increase in CN concentrations (at STP conditions, the relevant metric for comparisons between different altitudes) above the boundary layer. This implies high nucleation rates near the tropopause and that the upper troposphere is a net source of new particles for lower levels. Although nucleation events in the marine boundary layer have been observed [Clarke et al., 1998b], such observational evidence suggests that nucleation rates are higher, on average, in the free and upper troposphere.

7. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

[110] A size-resolved aerosol microphysical simulation has been incorporated into the GISS GCM II-prime general circulation model as an online component. The microphysical algorithm includes condensation/evaporation, coagulation, and nucleation. A two-moment sectional algorithm has been selected to represent aerosol microphysics. This algorithm has the desirable property of explicitly predicting and conserving both aerosol mass and number. Using two moments for each size section is numerically more accurate and efficient than single moment methods that track only aerosol mass. Moreover, conservation of aerosol number is an especially useful feature for predicting CCN number concentrations. The algorithm has high size resolution with 30 size sections between 0.01 and 10 μm diameter. It also represents a significant step toward including the indirect effect of aerosols in climate models in a mechanistic, rather than empirical, fashion.

[111] As a first application of the algorithm, the size distribution, mass, and number concentrations of sulfate aerosol have been predicted. Although, clearly there are many situations in which one would like to include other aerosol types, this is left for future work. Mass concentrations of sulfate predicted by the size-resolved model are essentially unchanged from previous model versions that lacked size resolution. The model also qualitatively reproduces several important features of the atmospheric aerosol. These include a large nucleation source in the upper troposphere, especially in the tropics, CN concentrations that increase with altitude, higher number concentrations over land than sea, and bimodal size distributions in marine areas that result from cloud processing. Quantitatively, predicted CCN concentrations are also within the expected range for both continental and marine locations. More detailed comparisons with observations show good agreement for size distributions, CN and CCN concentrations. Predicted CN concentrations are within 25% of observations although predictions may be biased low. Other discrepancies between observations and predictions, such as low CCN concentrations in the North Atlantic, may be explained by the absence of sea salt and carbonaceous aerosols.

[112] Sensitivity scenarios performed as part of this study make it clear that there are significant uncertainties with respect to sources of particles to the atmosphere. Emissions of primary particles to the atmosphere during fossil fuel combustion make the dominant contribution to CN concentrations in the polluted boundary layer and a have a noticeable impact in all but the most remote marine locations. It is clear that these need to be accurately included in global models that predict CN and CCN concentrations. Doing so will require more precise estimates of particulate emissions, including characterizing their size distributions. As a rough indication of the uncertainty associated with nucleation, a sensitivity scenario with a changed parameterization resulted in rates of new particle formation that were three times higher than the base case. Impact on CCN concentrations, however, was smaller but still significant, on the order of 25%. Comparison with observations suggested that the lower nucleation rates from the base case scenario, based on binary nucleation theory, generally resulted in the best predictions of CN concentrations. While this was true in the free and upper troposphere where nucleation has the largest impact, slightly higher nucleation rates improve model performance in the marine boundary layer. In contrast to the important sensitivities to surface-based primary emissions and nucleation, sensitivity scenarios exploring the effects of primary emissions from aircraft and scavenging of interstitial aerosol showed that the model is much less sensitive to these processes. A preliminary simulation of the preindustrial sulfate aerosol confirms that anthropogenic emissions have likely increased CCN concentrations significantly, but more conclusive results will await inclusion of other aerosol types.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References

[113] The authors would like to thank Graham Feingold for answering questions regarding the microphysics algorithm and providing sample code. Bernd Kärcher provided observational data in advance of publication. We thank Tim Bates and others at PMEL for making data sets from cruises available. Credit should go to two anonymous reviewers who took time to make thorough, specific, and constructive comments on a lengthy manuscript. This study has been supported by a graduate fellowship from the Fannie and John Hertz Foundation as well as by the National Aeronautics and Space Administration Earth Observing System Interdisciplinary Science program (NASA EOS-IDS).

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Formulation of Aerosol Microphysics
  5. 3. Coupling Aerosol Microphysics to the GCM
  6. 4. Simulated Microphysics of Tropospheric Sulfate
  7. 5. Sensitivity Scenarios
  8. 6. Comparison With Observations
  9. 7. Conclusions
  10. Acknowledgments
  11. References
  • Adams, P. J., J. H. Seinfeld, and D. M. Koch, Global concentrations of tropospheric sulfate, nitrate, and ammonium aerosol simulated in a general circulation model, J. Geophys. Res., 104, 13,79113,823, 1999.
  • Adams, P. J., J. H. Seinfeld, D. Koch, L. Mickley, and D. Jacob, General circulation model assessment of direct radiative forcing by the sulfate-nitrate-ammonium-water inorganic aerosol system, J. Geophys. Res., 106, 10971111, 2001.
  • Albrecht, B. A., Aerosols, cloud microphysics, and fractional cloudiness, Science, 245, 12271230, 1989.
  • Anderson, B. E., et al., An assessment of aircraft as a source of particles to the upper troposphere, Geophys. Res. Lett., 26, 30693072, 1999.
  • Barth, M. C., P. J. Rasch, J. T. Kiehl, C. M. Benkovitz, and S. E. Schwartz, Sulfur chemistry in the National Center for Atmospheric Research Community Climate Model: Description, evaluation, features, and sensitivity to aqueous chemistry, J. Geophys. Res., 105, 13871415, 2000.
  • Bates, T. S., V. N. Kasputin, P. K. Quinn, D. S. Covert, D. J. Coffman, C. Mari, P. A. Durkee, W. DeBruyn, and E. Saltzman, Processes controlling the distribution of aerosol particles in the lower marine boundary layer during the First Aerosol Characterization Experiment (ACE 1), J. Geophys. Res., 103, 16,36916,384, 1998.
  • Baughcum, S. L., D. M. Chan, S. M. Happenny, S. C. Henderson, P. S. Hertel, T. Higman, D. R. Maggiora, and C. A. Oncina, Emissions scenarios development: Scheduled 1990 and projected 2015 subsonic, Mach 2.0 and Mach 2.4 aircraft, in The Atmospheric Effects of Stratospheric Aircraft: A Third Program Report, NASA Ref. Publ., 1313, 89131, Nov. 1993.
  • Benkovitz, C. M., M. T. Scholtz, J. Pacyna, L. Tarrason, J. Dignon, E. C. Voldner, P. A. Spiro, J. A. Logan, and T. E. Graedel, Global gridded inventories of anthropogenic emissions of sulfur and nitrogen, J. Geophys. Res., 101, 29,23929,253, 1996.
  • Bigg, E. K., J. L. Gras, and C. Evans, Origin of Aitken particles in remote regions of the southern hemisphere, J. Atmos. Chem., 1, 203214, 1984.
  • Binkowski, F. S., and U. Shankar, The regional particulate matter model, 1, Model description and preliminary results, J. Geophys. Res., 100, 26,19126,209, 1995.
  • Boucher, O., GCM estimate of the indirect aerosol forcing using satellite-retrieved cloud droplet effective radii, J. Clim., 8, 14031409, 1995.
  • Boucher, O., and U. Lohmann, The sulfate-CCN-cloud albedo effect, Tellus, Ser. B, 47, 281300, 1995.
  • Brasseur, G. P., R. A. Cox, D. Hauglustaine, I. Isaksen, J. Lelieveld, D. H. Lister, R. Sausen, U. Schumann, A. Wahner, and P. Wiesen, European scientific assessment of the atmospheric effects of aircraft emissions, Atmos. Environ., 32, 23292418, 1998.
  • Brown, R. C., R. C. MiakeLye, M. R. Anderson, C. E. Kolb, and T. J. Resch, Aerosol dynamics in near-field aircraft plumes, J. Geophys. Res., 101, 22,93922,953, 1996.
  • Chin, M., D. J. Jacob, G. M. Gardner, M. S. Foreman-Fowler, and P. A. Spiro, A global three-dimensional model of tropospheric sulfate, J. Geophys. Res., 101, 18,66718,690, 1996.
  • Chuang, C. C., and J. E. Penner, Effects of anthropogenic sulfate on cloud drop nucleation and optical properties, Tellus, Ser. B, 47, 566577, 1995.
  • Chuang, C. C., J. E. Penner, K. E. Taylor, A. S. Grossman, and J. J. Walton, An assessment of the radiative effects of anthropogenic sulfate, J. Geophys. Res., 102, 37613778, 1997.
  • Clarke, A. D., Atmospheric nuclei in the remote free-troposphere, J. Atmos. Chem., 14, 479488, 1992.
  • Clarke, A. D., Atmospheric nuclei in the Pacific midtroposphere: Their nature, concentration, and evolution, J. Geophys. Res., 98, 20,63320,647, 1993.
  • Clarke, A. D., J. L. Varner, F. Eisele, R. L. Mauldin, D. Tanner, and M. Litchy, Particle production in the remote marine atmosphere: Cloud outflow and subsidence during ACE 1, J. Geophys. Res., 103, 16,39716,409, 1998a.
  • Clarke, A. D., et al., Particle nucleation in the tropical boundary layer and its coupling to marine sulfur sources, Science, 282, 8992, 1998b.
  • Clarke, A. D., F. Eisele, V. N. Kasputin, K. Moore, D. Tanner, L. Mauldin, M. Litchy, B. Lienert, M. A. Carroll, and G. Albercook, Nucleation in the equatorial free troposphere: Favorable environments during PEM-Tropics, J. Geophys. Res., 104, 57355744, 1999a.
  • Clarke, A. D., V. N. Kasputin, F. L. Eisele, R. J. Weber, and P. H. McMurry, Particle production near marine clouds: Sulfuric acid and predictions from classical binary nucleation, Geophys. Res. Lett., 26, 24252428, 1999b.
  • Coffman, D. J., and D. A. Hegg, A preliminary study of the effect of ammonia on particle nucleation in the marine boundary layer, J. Geophys. Res., 100, 71477160, 1995.
  • Cooke, W. F., and J. J. N. Wilson, A global black carbon aerosol model, J. Geophys. Res., 101, 19,39519,409, 1996.
  • Covert, D. S., V. N. Kasputin, T. S. Bates, and P. K. Quinn, Physical properties of marine boundary layer aerosol particles of the mid-Pacific in relation to sources and meteorological transport, J. Geophys. Res., 101, 69196930, 1996.
  • Dahneke, B., Simple kinetic theory of Brownian diffusion in vapors and aerosols, in Theory of Dispersed Multiphase Flow, edited by R. E. Meyer, pp. 97138, Academic, San Diego, Calif., 1983.
  • Dana, M. T., and J. M. Hales, Statistical aspects of the washout of polydisperse aerosols, Atmos. Environ., 10, 4550, 1976.
  • Del Genio, A. D., and M.-S. Yao, Efficient cumulus parameterization for long-term climate studies: The GISS scheme, in The Representation of Cumulus Convection in Numerical Models, Monogr. 46, edited by K. A. Emanuel, and D. J. Raymond, pp. 181184, Am. Meteorol. Soc., Boston, Mass., 1993.
  • Del Genio, A., D. Yao, M.-S. Kovari, and W. Lo, A prognostic cloud water parameterization for global climate models, J. Clim., 9, 270304, 1996.
  • Dentener, F. J., G. R. Carmichael, Y. Zhang, J. Lelieveld, and P. J. Crutzen, Role of mineral aerosol as a reactive surface in the global troposphere, J. Geophys. Res., 101, 22,86922,889, 1996.
  • Dinger, J. E., H. B. Howell, and T. A. Wojciechowski, On the source and composition of cloud condensation nuclei in the subsident air masses over the North Atlantic, J. Atmos. Sci., 27, 791797, 1970.
  • Doyle, G. J., Self-nucleation in the sulfuric acid-water system, J. Chem. Phys., 35, 795799, 1961.
  • Easter, R. C., and L. K. Peters, Binary homogeneous nucleation – Temperature and relative humidity fluctuations, nonlinearity, and aspects of new particle production in the atmosphere, J. Appl. Meteorol., 33, 775784, 1994.
  • Erickson, D. J., R. J. Oglesby, and S. Marshall, Climate response to indirect anthropogenic sulfate forcing, Geophys. Res. Lett., 22, 20172020, 1995.
  • Fahey, D. W., et al., Emission measurements of the Concorde supersonic aircraft in the lower stratosphere, Science, 270, 7073, 1995.
  • Feichter, J., E. Kjellstrom, H. Rodhe, F. Dentener, J. Lelieveld, and G.-J. Roelofs, Simulation of the tropospheric sulfur cycle in a global climate model, Atmos. Environ., 30, 16931707, 1996.
  • Feichter, J., U. Lohmann, and I. Schult, The atmospheric sulfur cycle in ECHAM-4 and its impact on the shortwave radiation, Clim. Dyn., 13, 235246, 1997.
  • Fitzgerald, J., Marine aerosols: A review, Atmos. Environ., 25, 533545, 1991.
  • Gallagher, M. W., K. M. Beswick, J. Duyzer, H. Westrate, T. W. Choularton, and P. Hummelshøj, Measurements of aerosol fluxes to Speulder forest using a micrometeorological technique, Atmos. Environ., 31, 359373, 1997.
  • Ghan, S. J., R. C. Easter, E. G. Chapman, H. Abdul-Razzak, Y. Zhang, L. R. Leung, N. S. Laulainen, R. D. Saylor, and R. A. Zaveri, A physically based estimate of radiative forcing by anthropogenic sulfate aerosol, J. Geophys. Res., 106, 52795293, 2001a.
  • Ghan, S., R. Easter, J. Hudson, and F.-M. Bréon, Evaluation of aerosol indirect radiative forcing in MIRAGE, J. Geophys. Res., 106, 53175334, 2001b.
  • Gong, S. L., L. A. Barrie, J.-P. Blanchet, and L. Spacek, Modeling size-distributed sea salt aerosols in the atmosphere: An application using Canadian climate models, in Air Pollution Modeling and Its Applications, vol. 12, edited by S.-E. Gryning, and N. Chaumerliac, Plenum, New York, 1998.
  • Hansen, J., G. Russell, D. Rind, P. Stone, A. Lacis, S. Lebedeff, R. Ruedy, and L. Travis, Efficient three-dimensional global models for climate studies: Models I and II. Mon. Weather Rev., 111, 609662, 1983.
  • Heist, R. H., and H. Reiss, Hydrates in supersaturated sulfuric acid-water vapor, J. Chem. Phys., 61, 573581, 1974.
  • Hoppel, W. A., J. E. Dinger, and R. E. Ruskin, Vertical profiles of CCN at various geographical locations, J. Atmos. Sci., 30, 14101419, 1973.
  • Hoppel, W. A., J. W. Fitzgerald, and R. E. Larson, Aerosol size distributions in air masses advecting off the East Coast of the United States, J. Geophys. Res., 90, 23652379, 1985.
  • Hoppel, W. A., G. M. Frick, and R. E. Larson, Effect of nonprecipitating clouds on the aerosol size distribution in the marine boundary layer, Geophys. Res. Lett., 13, 125128, 1986.
  • Intergovernmental Panel on Climate Change (IPCC), Third Assessment Report, Cambridge Univ. Press, New York, 2001.
  • Jaecker-Voirol, A., and P. Mirabel, Nucleation rate in a binary mixture of sulfuric acid and water vapor, J. Phys. Chem., 92, 35183521, 1988.
  • Jaecker-Voirol, A., and P. Mirabel, Heteromolecular nucleation in the sulfuric acid-water system, Atmos. Environ., 23, 20532057, 1989.
  • Jaenicke, R., Tropospheric aerosols, in Aerosol-Cloud-Climate Interactions, edited by P. V. Hobbs, pp. 131, Academic, San Diego, Calif., 1993.
  • Kanakidou, M., K. Tsigaridis, F. J. Dentener, and P. J. Crutzen, Human-activity-enhanced formation of organic aerosols by biogenic hydrocarbon oxidation, J. Geophys. Res., 105, 92439254, 2000.
  • Kasibhatla, P., W. L. Chameides, and J. St. John, A three-dimensional global model investigation of seasonal variations in the atmospheric burden of anthropogenic aerosols, J. Geophys. Res., 102, 37373759, 1997.
  • Kettle, A. J., and M. O. Andreae, Flux of dimethylsulfide from the oceans: A comparison of updated data sets and flux models, J. Geophys. Res., 105, 26,79326,808, 2000.
  • Kiehl, J. T., T. L. Schneider, P. J. Rasch, M. C. Barth, and J. Wong, Radiative forcing due to sulfate aerosols from simulations with the National Center for Atmospheric Research Community Climate Model, Version 3, J. Geophys. Res., 105, 14411457, 2000.
  • Kjellstrom, E., A three-dimensional global model study of carbonyl sulfide in the troposphere and the lower stratosphere, J. Atmos. Chem., 29, 151177, 1998.
  • Kjellstrom, E., J. Feichter, R. Sausen, and R. Hein, The contribution of aircraft emissions to the atmospheric sulfur budget, Atmos. Environ., 33, 34553465, 1999.
  • Koch, D., D. Jacob, I. Tegen, D. Rind, and M. Chin, Tropospheric sulfur simulation and sulfate direct radiative forcing in the Goddard Institute for Space Studies general circulation model, J. Geophys. Res., 104, 23,79923,822, 1999.
  • Korhonen, P., M. Kulmala, A. Laaksonen, Y. Viisanen, R. McGraw, and J. H. Seinfeld, Ternary nucleation of H2SO4, NH3, and H2O in the atmosphere, J. Geophys. Res., 104, 26,34926,353, 1999.
  • Kulmala, M., and A. Laaksonen, Binary nucleation of water-sulfuric acid system: Comparison of classical theories with different H2SO4 saturation vapor pressures, J. Chem. Phys., 93, 696701, 1990.
  • Kulmala, M., U. Pirjola, and J. M. Makela, Stable sulphate clusters as a source of new atmospheric particles, Nature, 404, 6669, 2000.
  • Laaksonen, A., V. Talanquer, and D. W. Oxtoby, Nucleation: Measurements, theory, and atmospheric applications, Annu. Rev. Phys. Chem., 46, 489524, 1995.
  • Langner, J., and H. Rodhe, A global three-dimensional model of the tropospheric sulfur cycle, J. Atmos. Chem., 13, 225263, 1991.
  • Lelieveld, J., G.-J. Roelofs, L. Ganzeveld, J. Feichter, and H. Rodhe, Terrestrial sources and distribution of atmospheric sulfur, Philos. Trans. R. Soc. London, 352, 149158, 1997.
  • Liousse, C., J. E. Penner, C. Chuang, J. J. Walton, H. Eddleman, and H. Cachier, A global three-dimensional model study of carbonaceous aerosols, J. Geophys. Res., 101, 19,41119,432, 1996.
  • Liu, X. H., D. A. Hegg, and M. T. Stoelinga, Numerical simulation of new particle formation over the northwest Atlantic using the MM5 mesoscale model coupled with sulfur chemistry, J. Geophys. Res., 106, 96979715, 2001.
  • Lohmann, U., and J. Feichter, Impact of sulfate aerosols on albedo and lifetime of clouds: A sensitivity study with the ECHAM4 GCM, J. Geophys. Res., 102, 13,68513,700, 1997.
  • Marti, J. J., R. J. Weber, P. H. McMurry, F. Eisele, D. Tanner, and A. Jefferson, New particle formation at a remote continental site: Assessing the contributions of SO2 and organic precursors, J. Geophys. Res., 102, 63316339, 1997.
  • Mirabel, P., and J. L. Katz, Binary homogeneous nucleation as a mechanism for the formation of aerosols, J. Chem. Phys., 60, 11381144, 1974.
  • Murphy, D. M., J. R. Anderson, P. K. Quinn, L. M. McInnes, F. J. Brechtel, S. M. Kreidenweis, A. M. Middlebrook, M. Pósfai, D. S. Thomson, and P. R. Buseck, Influence of sea-salt on aerosol radiative properties in the Southern Ocean marine boundary layer, Nature, 392, 6265, 1998.
  • Nenes, A., C. Pilinis, and S. N. Pandis, Isorropia: A new thermodynamic equilibrium model for multiphase multicomponent inorganic aerosols, Aquat. Geochem., 4, 123152, 1998.
  • Nilsson, E. D., and M. Kulmala, The potential for atmospheric mixing processes to enhance the binary nucleation rate, J. Geophys. Res., 103, 13811389, 1998.
  • Nilsson, E. D., L. Pirjola, and M. Kulmala, The effect of atmospheric waves on aerosol nucleation and size distribution, J. Geophys. Res., 105, 19,91719,926, 2000.
  • O'Dowd, C., M. H. Smith, I. E. Consterdine, and J. A. Lowe, Marine aerosol, sea-salt, and the marine sulphur cycle: A short review, Atmos. Environ., 31, 7380, 1997.
  • Penner, J. E., C. S. Atherton, and T. E. Graedel, Global emissions and models of photochemically active compounds, in Global Atmospheric-Biospheric Chemistry, edited by R. G. Prinn, pp. 223247, Plenum, New York, 1994.
  • Petzold, A., and F. P. Schröder, Jet engine exhaust aerosol characterization, Aerosol Sci. Technol., 28, 6276, 1998.
  • Petzold, A., A. Dopelheuer, C. A. Brock, and F. Schröder, In situ observations and model calculations of black carbon emission by aircraft at cruise altitude, J. Geophys. Res., 104, 22,17122,181, 1999.
  • Pham, M., J.-F. Müller, G. P. Brasseur, C. Granier, and G. Mégie, A three-dimensional study of the tropospheric sulfur cycle, J. Geophys. Res., 100, 26,06126,092, 1995.
  • Pincus, R., and M. Baker, Effect of precipitation on the albedo susceptibility of clouds in the marine boundary layer, Nature, 372, 250252, 1994.
  • Pöschl, U., M. Canagaratna, J. T. Jayne, L. T. Molina, D. R. Worsnop, C. E. Kolb, and M. J. Molina, Mass accommodation coefficient of H2SO4 vapor on aqueous sulfuric acid surfaces and gaseous diffusion coefficient of H2SO4 in N2/H2O, J. Phys. Chem. A, 102, 10,08210,089, 1998.
  • Quinn, P. K., S. F. Marshall, T. S. Bates, D. S. Covert, and V. N. Kasputin, Comparison of measured and calculated aerosol properties relevant to the direct radiative forcing of tropospheric sulfate aerosol on climate, J. Geophys. Res., 100, 89778991, 1995.
  • Raes, F., J. Wilson, and R. Van Dingenen, Aerosol dynamics and its implication for the global aerosol climatology, in Aerosol Forcing of Climate, edited by R. J. Charlson, and J. Heintzenberg, pp. 153169, John Wiley, New York, 1995.
  • Raes, F., R. Van Dingenen, E. Cuevas, P. F. J. Van Velthoven, and J. M. Prospero, Observations of aerosols in the free troposphere and marine boundary layer of the subtropical northeast Atlantic: Discussion of processes determining their size distribution, J. Geophys. Res., 102, 21,31521,328, 1997.
  • Rasch, P. J., M. C. Barth, J. T. Kiehl, S. E. Schwartz, and C. M. Benkovitz, A description of the global sulfur cycle and its controlling processes in the National Center for Atmospheric Research Community Climate Model, Version 3, J. Geophys. Res., 105, 13671385, 2000.
  • Restad, K., I. S. A. Isaksen, and T. K. Berntsen, Global distribution of sulphate in the troposphere: A three-dimensional model study, Atmos. Environ., 32, 35933609, 1998.
  • Rind, D., and J. Lerner, The use of on-line tracers as a diagnostic tool in general circulation model development, 1, Horizontal and vertical transport in the troposphere, J. Geophys. Res., 101, 12,66712,683, 1996.
  • Roelofs, G.-J., J. Lelieveld, and L. Ganzeveld, Simulation of global sulfate distribution and the influence on effective cloud drop radii with a coupled photochemistry-sulfur cycle model, Tellus, Ser. B, 50, 224242, 1998.
  • Russell, L. M., and J. H. Seinfeld, Size- and composition-resolved externally mixed aerosol model, Aerosol Sci. Technol., 28, 403416, 1998.
  • Schröder, F. P., B. Kärcher, A. Petzold, and M. Fiebig, Aerosol sources and transformation processes in the free troposphere and tropopause region at northern midlatitudes, J. Geophys. Res., 107, doi:10.1029/2000JD000194, in press, 2002.
  • Schumann, U., J. Strom, R. Busen, R. Baumann, K. Gierens, M. Krautstrunk, F. P. Schröder, and J. Stingl, In situ observations of particles in jet aircraft exhausts and contrails for different sulfur-containing fuels, J. Geophys. Res., 101, 68536869, 1996.
  • Sehmel, G. A., Particle and gas deposition: A review, Atmos. Environ., 14, 9831011, 1980.
  • Seinfeld, J. H., and S. N. Pandis, Atmospheric Chemistry and Physics, John Wiley, New York, 1998.
  • Takemura, T., H. Okamoto, Y. Maruyama, A. Numaguti, A. Higurashi, and T. Nakajima, Global three-dimensional simulation of aerosol optical thickness distribution of various origins, J. Geophys. Res., 105, 17,85317,873, 2000.
  • Tang, I. N., and H. R. Munkelwitz, Water activities, densities, and refractive indices of aqueous sulfates and sodium nitrate droplets of atmospheric importance, J. Geophys. Res., 99, 18,80118,808, 1994.
  • Tegen, I., and I. Fung, Modeling of mineral dust in the atmosphere: Sources, transport, and optical thickness, J. Geophys. Res., 99, 22,89722,914, 1994.
  • Tegen, I., and A. A. Lacis, Modeling of particle size distribution and its influence on the radiative properties of mineral dust aerosol, J. Geophys. Res., 101, 19,23719,244, 1996.
  • Tegen, I., P. Hollrig, M. Chin, I. Fung, D. Jacob, and J. Penner, Contribution of different aerosol species to the global aerosol extinction optical thickness: Estimates from model results, J. Geophys. Res., 102, 23,89523,915, 1997.
  • Tegen, I., D. Koch, A. A. Lacis, and M. Sato, Trends in tropospheric aerosol loads and corresponding impact on direct radiative forcing between 1950 and 1990: A model study, J. Geophys. Res., 105, 26,97126,989, 2000.
  • Turco, R. P., J.-X. Zhao, and F. Yu, A new source of tropospheric aerosols: Ion-ion recombination, Geophys. Res. Lett., 25, 635638, 1998.
  • Twomey, S., Pollution and the planetary albedo, Atmos. Environ., 8, 12511256, 1974.
  • Tzivion, S., G. Feingold, and Z. Levin, An efficient numerical solution to the stochastic collection equation, J. Atmos. Sci., 44, 31393149, 1987.
  • Tzivion, S., G. Feingold, and Z. Levin, The evolution of raindrop spectra, part II, Collisional collection/breakup and evaporation in a rainshaft, J. Atmos. Sci., 46, 33123327, 1989.
  • Van Dingenen, R., F. Raes, and N. R. Jensen, Evidence for anthropogenic impact on number concentration and sulfate content of cloud-processed aerosol particles over the North Atlantic, J. Geophys. Res., 100, 21,05721,067, 1995.
  • von Salzen, K., H. G. Leighton, P. A. Ariya, L. A. Barrie, S. L. Gong, J. P. Blanchet, L. Spacek, U. Lohmann, and L. I. Kleinman, Sensitivity of sulphate aerosol size distributions and CCN concentrations over North America to SOx emissions and H2O2 concentrations, J. Geophys. Res., 105, 97419765, 2000.
  • Wesely, M. L., and B. B. Hicks, Some factors that affect the deposition rates of sulfur dioxide and similar gases on vegetation, J. Air Pollut. Control Assoc., 27, 11101116, 1977.
  • Wexler, A. S., F. W. Lurmann, and J. H. Seinfeld, Modeling urban and regional aerosols, I, Model development, Atmos. Environ., 28, 531546, 1994.
  • Whitby, K. T., The physical characteristics of sulfur aerosols, Atmos. Environ., 12, 135159, 1978.
  • Zhang, Y. P., S. Kreidenweis, and G. R. Taylor, The effects of clouds on aerosol and chemical species production and distribution, part III, Aerosol model description and sensitivity analysis, J. Atmos. Sci., 55, 921939, 1998.