《环境物理化学》教学案例：吸湿增长与表面现象 The effect of physical and chemical aerosol properties on warm cloud droplet activation

2598 G. McFiggans et al. aerosol effects on warm cloud activation fully soluble particles of greater than 40 nm diameter may 0.950.960.970.980.991.00 1400 activate under certain atmospheric circumstances. Studies of low clouds have found minimum activated diameters ranging from 40 to 140 nm( Hoppel et al., 1996; Leaitch et al., 1996 Flynn et al., 2000; Komppula et al., 2005). Smaller particles might be activated in the strong updrafts of vigorous convec- tion, especially in unpolluted environments. The activation E of particles above this dry size threshold is a strong function of both the supersaturation and the form of the size and com- position distribution. A primary limitation of the approach of Twomey is that it does not account for the multi-component size and composition of the CCn population 3.1.2 A practical illustration of the dynamic competition effects of a simple multi-component aerosol popula- 1000 0.01 0.10 1.00 The Twomey equation, and most others developed to link radius, um number of activated droplets to a sub-cloud aerosol con- centration(either mass or number), usually assume that the Fig 4. Simulation showing the change in droplet radius with height aerosol type is relatively homogeneous. Some parameterisa- in a simulation initialised with an ammonium sulphate aerosol with tions yield a monotonic increase in cloud droplet concentra- a geometric mean diameter of 140 nm, a geometric standard de- viation, g of 1.7 and aerosol number concentration of 300cm<3 tion(e. g, Boucher and Lohmann, 1995)whereas others show (corresponding to a total mass loading of 0. 76 ugm ). The simu- 1994). In general they are well approximated by power ation was started at an RH of% at 1000 m. Solid lines represent law dependence of Na on Na or the mass concentration of clected aerosol size classes. The dashed line is the saturation ratio aerosols. The parameterisations of Ghan et al.(1998)and Feingold (2001), exhibit similar behaviour but, following sat- uration, show a decrease in Nd with further increases with Na nuclei activated at peak supersaturation $, indicates the at very high aerosol concentrations(order 10000 cm relative sensitivity of activated drop number to aerosol size distribution parameters. A(T, P)is a function of temper- aerosol but not for an aerosol of significantly different multi- ature T and pressure P and B is the complete beta func le components. The breakdown of these assumptions was tion. This expression works well for modest values of k well illustrated by Ghan et al. (1998)and O'Dowd et al and was evaluated for southern-hemisphere cumulus clouds. ( 1999) for typical maritime cloud conditions Although the analysis assumes a Junge power-law distribu- O Dowd et al. (1999)conducted parcel model simulations tion, the essential features of sensitivity to size parameters of the impact of increasing CCN concentrations on cloud and updraught is evident. For example, drop activation is in- droplet concentration using typical marine sulphate and sea- creasingly more sensitive to updraught velocity when size salt size distributions and found that an increase in cloud distributions have larger values of k(steeper decreases in droplet concentration did not necessarily follow as the CCN concentration with increasing size). Conversely, at smaller values of k, N is determined primarily by c(Twomey 7 Population increases. The aerosol properties were described by three modes, one sulphate mode with a modal diameter Jaenicke( 1988)pointed out that the power-law representa- of 150 nm and a geometric standard deviation, ag of 1.4,a tion of the CCN spectrum may not be realistic, particularly film drop sea-salt mode with modal diameter of 200 nm and for a multi-component aerosol population. Theoretical and og of 1. 9 and a super micron mode of 2 um and og of 2. The model-based analyses of activation in terms of lognormal amplitude of each mode was varied within the frequently ob- size distributions have been reported by Cohard et al. (1998): served constraints that the amplitude of the sulphate mode is Feingold (2001, 2003), Rissman et al. (2004) Abdul-Razzak typically greater than those of the sea-salt modes and the am- and ghan(2000) Nenes and Seinfeld(2003), Fountoukis plitude of the film drop mode is always significantly greater than that of the jet drop mode. Simulations were conducted tinsson et al 1999) for scenarios where the sulphate aerosol The updraught velocity can range from around 0. 1 ms- varied under a range of fixed sea-salt conditions.Sea-salt in stratiform clouds to in excess of 15 ms- in deep convec- CCN were added to the population with the base case of tive or orographically-forced clouds. The supersaturations 3 cm-3, increasing to a maximum of 57 cm-3 sea-salt par associated with such a range in updraught velocity mean that ticles tmos.Chem.Phys,6,2593-2649,2006 www.atmos-chem-phys.net/6/2593/2006/

2598 G. McFiggans et al.: Aerosol effects on warm cloud activation Fig. 4. Simulation showing the change in droplet radius with height in a simulation initialised with an ammonium sulphate aerosol with a geometric mean diameter of 140 nm, a geometric standard de￾viation, σg of 1.7 and aerosol number concentration of 300 cm−3 (corresponding to a total mass loading of 0.76 µgm−3 ). The simu￾lation was started at an RH of 95% at 1000 m. Solid lines represent selected aerosol size classes. The dashed line is the saturation ratio. of nuclei activated at peak supersaturation S ∗ , indicates the relative sensitivity of activated drop number to aerosol size distribution parameters. A(T , P ) is a function of temper￾ature T and pressure P and β is the complete beta func￾tion. This expression works well for modest values of k and was evaluated for southern-hemisphere cumulus clouds. Although the analysis assumes a Junge power-law distribu￾tion, the essential features of sensitivity to size parameters and updraught is evident. For example, drop activation is in￾creasingly more sensitive to updraught velocity when size distributions have larger values of k (steeper decreases in concentration with increasing size). Conversely, at smaller values of k, N is determined primarily by c (Twomey, 1977). Jaenicke (1988) pointed out that the power-law representa￾tion of the CCN spectrum may not be realistic, particularly for a multi-component aerosol population. Theoretical and model-based analyses of activation in terms of lognormal size distributions have been reported by Cohard et al. (1998); Feingold (2001, 2003); Rissman et al. (2004); Abdul-Razzak and Ghan (2000); Nenes and Seinfeld (2003); Fountoukis and Nenes (2005); and experimentally determined (e.g. Mar￾tinsson et al., 1999). The updraught velocity can range from around 0.1 ms−1 in stratiform clouds to in excess of 15 ms−1 in deep convec￾tive or orographically-forced clouds. The supersaturations associated with such a range in updraught velocity mean that fully soluble particles of greater than ∼40 nm diameter may activate under certain atmospheric circumstances. Studies of low clouds have found minimum activated diameters ranging from 40 to 140 nm (Hoppel et al., 1996; Leaitch et al., 1996; Flynn et al., 2000; Komppula et al., 2005). Smaller particles might be activated in the strong updrafts of vigorous convec￾tion, especially in unpolluted environments. The activation of particles above this dry size threshold is a strong function of both the supersaturation and the form of the size and com￾position distribution. A primary limitation of the approach of Twomey is that it does not account for the multi-component size and composition of the CCN population. 3.1.2 A practical illustration of the dynamic competition effects of a simple multi-component aerosol popula￾tion The Twomey equation, and most others developed to link number of activated droplets to a sub-cloud aerosol con￾centration (either mass or number), usually assume that the aerosol type is relatively homogeneous. Some parameterisa￾tions yield a monotonic increase in cloud droplet concentra￾tion (e.g., Boucher and Lohmann, 1995) whereas others show a tendency to saturate (for example Hegg, 1984; Jones et al., 1994). In general they are well approximated by power￾law dependence of Nd on Na or the mass concentration of aerosols. The parameterisations of Ghan et al. (1998) and Feingold (2001), exhibit similar behaviour but, following sat￾uration, show a decrease in Nd with further increases with Na at very high aerosol concentrations (order 10 000 cm−3 ). Such assumptions may hold for a single component aerosol but not for an aerosol of significantly different multi￾ple components. The breakdown of these assumptions was well illustrated by Ghan et al. (1998) and O’Dowd et al. (1999) for typical maritime cloud conditions. O’Dowd et al. (1999) conducted parcel model simulations of the impact of increasing CCN concentrations on cloud droplet concentration using typical marine sulphate and sea￾salt size distributions and found that an increase in cloud droplet concentration did not necessarily follow as the CCN population increases. The aerosol properties were described by three modes; one sulphate mode with a modal diameter of 150 nm and a geometric standard deviation, σg of 1.4; a film drop sea-salt mode with modal diameter of 200 nm and σg of 1.9 and a super micron mode of 2 µm and σg of 2. The amplitude of each mode was varied within the frequently ob￾served constraints that the amplitude of the sulphate mode is typically greater than those of the sea-salt modes and the am￾plitude of the film drop mode is always significantly greater than that of the jet drop mode. Simulations were conducted for scenarios where the sulphate aerosol concentration was varied under a range of fixed sea-salt conditions. Sea-salt CCN were added to the population with the base case of 3 cm−3 , increasing to a maximum of 57 cm−3 sea-salt par￾ticles. Atmos. Chem. Phys., 6, 2593–2649, 2006 www.atmos-chem-phys.net/6/2593/2006/

G. McFiggans et al. Aerosol effects on warm cloud activation 2599 300 200 100 draft=0. 1 draft=0. 175 m s 0 01002003004000100200300400 Sub-cloud aerosol (cm Fig. 5. Cloud droplet concentration as a function of sub-cloud aerosol where the sub-cloud aerosol comprises an extemal mix of sulphate nd sea-salt CCN The simulations results are shown in Fig. 5 for updraughts sation of these properties(e. g. Whitby, 1978; Van Dingenen of 0. I ms- and 0. 175 ms. For low sulphate CCN concen- et al., 2004) trations, the addition of sea-salt CCN increases the number of Figure 6 shows representative average distributions in a activated droplets significantly while for high sulphate con- variety of locations. Most particles greater than 200 nm di- centrations, the number of activated droplets decreases sig- ameter with moderate amounts of soluble material will acti- nificantly. For the higher updraught, the point at which the vate under reasonable supersaturations. Assuming that those result of the addition of sea-salt nuclei switches from an in- particles greater than 200 nm in Fig. 6 are moderately sol- crease to a decrease in droplet concentration reduces for the uble, it can be seen that the sizes critical to determining the gher updraught and the impact of the reduction in droplet droplet number in an aerosol population fall in the range with concentration increases for increasing updraught significant contributions from both Aitken and accumulation The main processes driving this phenomenon are(1)sea- mode particles(around 100 nm diameter). It is therefore nec- salt CCN are typically larger than sulphate CCN; (2)for a essary to capture the features of the aerosol size distribu- en size, sea-salt is more active as a CCn than sulphate, tion in both modes in order to realistically describe cloud ()although the relative concentration of larger sea-salt CCn activation behaviour, (Martinsson et al., 1999). The follow is significantly lower than sulphate CCN, they contribute to ing sections investigate further properties of real atmospheric a significant reduction in the peak supersaturation reached in aerosol and the potential impacts of these properties on cloud cloud and thus inhibit the activation of sulphate nuclei. This activation example demonstrates that for even simple two-component It should be noted that, the critical size range for cloud aerosol systems the dynamic competition is quite complex activation of about 50 to 150 nm is not accessible to most and non-linear and that the effect of increasing the availabil- optical sizing instruments, but may be probed by mobility ity of ccn does not necessarily lead to an increase in cloud instruments. This significantly reduces the amount of data droplet concentration. Similar non-linearities are evident in available at cloud altitudes because mobility analyses can be the effects of composition on droplet activation and caution challenging on aircraft due to the time required to scan a size should be exercised in translating a composition change to distribution an equivalent change in drop number concentration(Ervens et al., 2005). The results of such responses are strongly de- 3.1.3 Relative importance of size distribution, composition pendent on water vapour supply (i.e. updraught)and conden sation rates(dependent on size distribution and composition) The activation of seasalt and sulphate in marine stratiform Feingold (2003) performed a sensitivity anal cloud described in this section is a particularly simple case in aspects of the relative importance of aerosol size and compo- terms of both composition and the limited range of updraught sition, in so far as both properties affect activation, using a velocity. Ambient aerosol size distributions are highly vari- cloud parcel model. Input aerosol size distributions(parame- able from location to location. The reader is referred to a terised as lognormal functions described by Na, rg, g), and range of review articles for a broad and detailed characteri- prescribed updraught velocities, w, were varied over a large www.atmos-chem-phys.net/6/2593/2006/ Atmos. Chem. Phys., 6, 2593-2649, 2006

G. McFiggans et al.: Aerosol effects on warm cloud activation 2599 Fig. 5. Cloud droplet concentration as a function of sub-cloud aerosol where the sub-cloud aerosol comprises an external mix of sulphate and sea-salt CCN. The simulations results are shown in Fig. 5 for updraughts of 0.1 ms−1 and 0.175 ms−1 . For low sulphate CCN concen￾trations, the addition of sea-salt CCN increases the number of activated droplets significantly while for high sulphate con￾centrations, the number of activated droplets decreases sig￾nificantly. For the higher updraught, the point at which the result of the addition of sea-salt nuclei switches from an in￾crease to a decrease in droplet concentration reduces for the higher updraught and the impact of the reduction in droplet concentration increases for increasing updraught. The main processes driving this phenomenon are (1) sea￾salt CCN are typically larger than sulphate CCN; (2) for a given size, sea-salt is more active as a CCN than sulphate; (3) although the relative concentration of larger sea-salt CCN is significantly lower than sulphate CCN, they contribute to a significant reduction in the peak supersaturation reached in cloud and thus inhibit the activation of sulphate nuclei. This example demonstrates that for even simple two-component aerosol systems the dynamic competition is quite complex and non-linear and that the effect of increasing the availabil￾ity of CCN does not necessarily lead to an increase in cloud droplet concentration. Similar non-linearities are evident in the effects of composition on droplet activation and caution should be exercised in translating a composition change to an equivalent change in drop number concentration (Ervens et al., 2005). The results of such responses are strongly de￾pendent on water vapour supply (i.e. updraught) and conden￾sation rates (dependent on size distribution and composition). The activation of seasalt and sulphate in marine stratiform cloud described in this section is a particularly simple case in terms of both composition and the limited range of updraught velocity. Ambient aerosol size distributions are highly vari￾able from location to location. The reader is referred to a range of review articles for a broad and detailed characteri￾sation of these properties (e.g. Whitby, 1978; Van Dingenen et al., 2004). Figure 6 shows representative average distributions in a variety of locations. Most particles greater than 200 nm di￾ameter with moderate amounts of soluble material will acti￾vate under reasonable supersaturations. Assuming that those particles greater than 200 nm in Fig. 6 are moderately sol￾uble, it can be seen that the sizes critical to determining the droplet number in an aerosol population fall in the range with significant contributions from both Aitken and accumulation mode particles (around 100 nm diameter). It is therefore nec￾essary to capture the features of the aerosol size distribu￾tion in both modes in order to realistically describe cloud activation behaviour, (Martinsson et al., 1999). The follow￾ing sections investigate further properties of real atmospheric aerosol and the potential impacts of these properties on cloud activation. It should be noted that, the critical size range for cloud activation of about 50 to 150 nm is not accessible to most optical sizing instruments, but may be probed by mobility instruments. This significantly reduces the amount of data available at cloud altitudes because mobility analyses can be challenging on aircraft due to the time required to scan a size distribution. 3.1.3 Relative importance of size distribution, composition and updraught Feingold (2003) performed a sensitivity analysis comparing aspects of the relative importance of aerosol size and compo￾sition, in so far as both properties affect activation, using a cloud parcel model. Input aerosol size distributions (parame￾terised as lognormal functions described by Na, rg, σg), and prescribed updraught velocities, w, were varied over a large www.atmos-chem-phys.net/6/2593/2006/ Atmos. Chem. Phys., 6, 2593–2649, 2006

2600 G. McFiggans et al. aerosol effects on warm cloud activation Background Near-city Urban marylebone(GB) 0oo10010.1 1E·5 E+5 E+5 Mept (Dj 010101 Jungtaujoch (CH pra(n 0o010011 Fig. 6. Median particle number size distributions during summer, during morning hours(black dashed line), afternoon(grey full line) and night(black full line). From Van Dingenen et al. (2004) Table 1. Table of s(Xi)=aIn Na/aln Xi where Xi is one of Na mately the same(although opposite in sign) whereas S(w is small. Under polluted conditions, the relative influence indicates Na>1000 cm-3. The ranges of x are w: 20 cms-I to of rg, Og and w on Na increases significantly while S(Na) 300, Na: 20 cm-3 to 3000 cm-3, rg: 0.03 to 0. 1 um, a: 1.3 decreases in importance. S(e) is relatively small compared to2.2,E:0.10to1.00 to the other terms. although we caution that this term only reflects composition changes associated with the fraction all Clean polluted of soluble material. The signs of S(Xi) are as expected specific mention is made of S(og) which is negative because Na0.880.92 0.7 rg0.320.280.39 the tail of the distribution at large sizes results in activation a-0.39-0.310.53 of larger drops, and suppression of supersaturation which tends to suppress Nd. This combination of effects makes E0.110.09 0.13 S(og) quite large, particularly under polluted conditions when the larger particles are abundant (e.g. O Dowd et al 1999, Sect. 3.1.2). Rissman et al.(2004)performed a more detailed analysis of the effect of various composition ange of parameter space. Aerosol composition was repre factors such as solubility and surface tension, as well as size sented in a simplified fashion by considering an ammonium distribution parameters. Their results were derived from sulphate and insoluble mix, and varying the mass fraction analytical solutions, and presented in terms of a sensitivity of ammonium sulphate, over the range 0. 1 to 1.0. The out- relative to the sensitivity of drop number concentration to put was then used to examine the relative sensitivity of cloud updraft velocity (x)=(x/w)(aNa/a x)/(aNa/aw),where drop size to the various input parameters using the model of x is a composition factor such as organic mass fraction Feingold(2003). Here we repeat this analysis for sensitiv- Eo. The authors concluded that when defined this way, ity of drop number concentration Na. The sensitivities defined as S(Xi)=aIn Nd/aIn Xi where Xi is one of Na,r sensitivity to composition factors (x) is highest for aerosol e). In this form, values of S(Xi) can be compared updraught velocity. However, these are conditions under with one-another to assess their relative importance. Values which supersaturation and activated fractions of S(Xi) for conditions similar to Feingold (2003)are given an increase in w does not add many new drops(aNa/dw in Table 1 is small). The appea of aNd/aw in the denominator Under clean conditions, arbitrarily defined as tends to increase (x). Thus at high S, even though Na<1000 cm -', S(Na) is close to its theoretical upper o(x) is large, composition effects may not be important limit of 1, indicating a high level of in-cloud activation in an absolute sense. Because the individual sensitivities Sensitivity to rg and og under clean conditions is approxi tmos.Chem.Phys,6,2593-2649,2006 www.atmos-chem-phys.net/6/2593/2006/

2600 G. McFiggans et al.: Aerosol effects on warm cloud activation Fig. 6. Median particle number size distributions during summer, during morning hours (black dashed line), afternoon (grey full line) and night (black full line). From Van Dingenen et al. (2004); van Din￾genen et al. (2004). 125 Fig. 6. Median particle number size distributions during summer, during morning hours (black dashed line), afternoon (grey full line) and night (black full line). From Van Dingenen et al. (2004). Table 1. Table of S(Xi )=∂lnNd /∂lnXi where Xi is one of Na, rg, σg, w, ε. “Clean” indicates Na<1000 cm−3 and “Polluted” indicates Na>1000 cm−3 . The ranges of Xi are w: 20 cm s−1 to 300 cm s−1 , Na: 20 cm−3 to 3000 cm−3 , rg: 0.03 to 0.1µm, σ: 1.3 to 2.2, ε: 0.10 to 1.00. All Clean Polluted Na 0.88 0.92 0.73 rg 0.32 0.28 0.39 σ −0.39 −0.31 −0.53 w 0.29 0.18 0.47 ε 0.11 0.09 0.13 range of parameter space. Aerosol composition was repre￾sented in a simplified fashion by considering an ammonium sulphate and insoluble mix, and varying the mass fraction of ammonium sulphate, over the range 0.1 to 1.0. The out￾put was then used to examine the relative sensitivity of cloud drop size to the various input parameters using the model of Feingold (2003). Here we repeat this analysis for sensitiv￾ity of drop number concentration Nd . The sensitivities are defined as S(Xi)=∂lnN d/∂lnXi where Xi is one of Na, rg, σg, w or ε). In this form, values of S(Xi) can be compared with one-another to assess their relative importance. Values of S(Xi) for conditions similar to Feingold (2003) are given in Table 1. Under clean conditions, arbitrarily defined as Na<1000 cm−3 , S(Na) is close to its theoretical upper limit of 1, indicating a high level of in-cloud activation. Sensitivity to rg and σg under clean conditions is approxi￾mately the same (although opposite in sign) whereas S(w) is small. Under polluted conditions, the relative influence of rg, σg and w on Nd increases significantly while S(Na) decreases in importance. S(ε) is relatively small compared to the other terms, although we caution that this term only reflects composition changes associated with the fraction of soluble material. The signs of S(Xi) are as expected; specific mention is made of S(σg) which is negative because the tail of the distribution at large sizes results in activation of larger drops, and suppression of supersaturation which tends to suppress Nd . This combination of effects makes S(σg) quite large, particularly under polluted conditions when the larger particles are abundant (e.g. O’Dowd et al., 1999, Sect. 3.1.2). Rissman et al. (2004) performed a more detailed analysis of the effect of various composition factors such as solubility and surface tension, as well as size distribution parameters. Their results were derived from analytical solutions, and presented in terms of a sensitivity relative to the sensitivity of drop number concentration to updraft velocity φ(χ )=(χ/w)(∂Nd /∂χ )/(∂Nd /∂w), where χ is a composition factor such as organic mass fraction o. The authors concluded that when defined this way, sensitivity to composition factors φ(χ ) is highest for aerosol typical of marine condition, and increases with increasing updraught velocity. However, these are conditions under which supersaturation and activated fractions are high, and an increase in w does not add many new drops (∂Nd /∂w is small). The appearance of ∂Nd /∂w in the denominator tends to increase φ(χ ). Thus at high S, even though φ(χ ) is large, composition effects may not be important in an absolute sense,. Because the individual sensitivities Atmos. Chem. Phys., 6, 2593–2649, 2006 www.atmos-chem-phys.net/6/2593/2006/

G. McFiggans et al. Aerosol effects on warm cloud activation d/ax and aNa/aw or their logarithmic equivalents) Some online instruments provide single particle composition not reported, it is difficult to compare their results to information and hence information about which components those of Feingold(2003) for overlapping parameter space. co-exist in the same particles; their chemical mixing state Both studies do however agree that Nd is more sensitive to Others may provide component mass loadings with high size the size parameter rg, and that Nd is more sensitive to rg and time resolution. However, online techniques cannot cur- under polluted conditions rently provide as much detailed speciation information c The greater sensitivity of cloud droplet number to size may be available from offline techniques mpared to composition illustrates that the aerosol size must be captured as a primary pre-requisite. The sensitivity to the 3. 1. 4. I Offline bulk measurements showing the complexity compositional complexities should only be investigated in of the atmospheric aerosol composition e knowledge that the size and number information is likely to be equally important(or moreso). It should be noted that This section presents such features of aerosol composition the treatment of composition does not address the sensitivity (organic and inorganic)which may be gained from offline to composition changes with size and to varied composition analyses as relate to cloud activation, without attempting an at any one size; evidence for the prevalence of both being exhaustive review. A summary of some available size segre- provided in the forthcoming sections. The sensitivity of acti- gated chemical compositions is also provided in a form suit vation and cloud droplet number to more detailed aspects of able for cloud modelling purposes aerosol composition is discussed in Sect. 4.2 Despite the wide range of sampling and analytical tech- niques that have been employed, characterisation of aerosol 3.1. 4 The composition of ambient aerosol chemical composition as a function of size is often still in- Until recently, the vast majority of cloud modelling studies is diverse, complex and variable with location and condi- have conventionally assumed, implicitly or explicitly, that tion. The particles comprise many inorganic and organic the soluble material in aerosol particles comprises inorganic compounds ranging in solubility, density and, surface ten- components. The emerging complexity of ambient aerosol sion. Thus, comprehensive papers about the aspects of the equires that this description is revisited chemical composition relevant to cloud formation are rare It must be remembered that number of activated droplets is In order to use Eq.(2), the chemical composition must be dependent on the number distribution of particles of a given"translated"for cloud modelling purposes into concentra type and not directly on the mass loading. Although the acti- tion of molecules(organic and inorganic)dissolved in cloud vation of an individual particle is dependent on its(soluble) droplets, total insoluble mass and, if present, the concen- mass, techniques which coarsely probe component distribu- tration of some"critical" species with limited solubility(or tion loadings by mass will not provide adequate insight to slightly soluble species). The degree of dissociation and ph predict droplet number. Composition is likely to be impor- may also be needed (Laaksonen et al., 1998; Lohmann et al tant only in a limited size range: particles smaller than about 2004); this is addressed further in Sect. 4. It should be noted 40 nm diameter are unlikely to activate into cloud droplets that there are exceptions to these requirements when Eq (1) regardless of their composition and sufficiently large parti- is used, depending on how water activity is evaluated, see cles will almost always contain enough soluble material to Sect. 4.4 activate. To predict droplet activation it is necessary to de- Soluble inorganic components are relatively well un- termine size-resolved composition in the 40 to 200 nm size derstood; the majority comprising a few inorganic range coupled to information about the mixing-state of the salts (Heintzenberg, 1989), which are relatively well- population. No single technique can currently provide all characterised in terms of their hygroscopic properties( Clegg this information. This section reviews the available evidence et al., 1998; Ansari and Pandis, 2000). The insoluble inor- for the composition of ambient particles from a range of ganic fraction can also be important(as in the case of dust studies and techniques From combinations of these sources aerosol from urban or natural sources)and many different it should eventually be possible to adequately describe the component or element measurements are available. How- aerosol composition distribution for purposes of CCN and ever, this information is difficult to directly interpret in terms possibly droplet number prediction of total insoluble mass Offline analyses of bulk particulate material collected by Organic matter has been shown to represent an important filter pack and impactor sampling have conventionally been fraction of the aerosol mass in different environments, and is used to determine aerosol mass composition. Applied ana- routinely measured by means of thermal techniques liousse lytical techniques can provide information ranging from de- et al., 1996, Jacobson et al., 2000; Putaud et al., 2004) tailed molecular speciation to aggregate lumped chemical ganic carbon(OC)and Elemental carbon(EC)(or black car- functionality. These techniques have recently been supple- bon(BC)are reported in terms of carbon mass and the trans- mented by online instrumentation which may provide addi- formation to aerosol mass is problematic without knowing tional information to that available from offline techniques. the main chemical C structure( Turpin and Lim, 2001; Rus- www.atmos-chem-phys.net/6/2593/2006/ Atmos. Chem. Phys., 6, 2593-2649, 2006

G. McFiggans et al.: Aerosol effects on warm cloud activation 2601 (∂Nd /∂χ and ∂Nd /∂w or their logarithmic equivalents) were not reported, it is difficult to compare their results to those of Feingold (2003) for overlapping parameter space. Both studies do however agree that Nd is more sensitive to the size parameter rg, and that Nd is more sensitive to rg under polluted conditions. The greater sensitivity of cloud droplet number to size compared to composition illustrates that the aerosol size must be captured as a primary pre-requisite. The sensitivity to the compositional complexities should only be investigated in the knowledge that the size and number information is likely to be equally important (or moreso). It should be noted that the treatment of composition does not address the sensitivity to composition changes with size and to varied composition at any one size; evidence for the prevalence of both being provided in the forthcoming sections. The sensitivity of acti￾vation and cloud droplet number to more detailed aspects of aerosol composition is discussed in Sect. 4.2. 3.1.4 The composition of ambient aerosol Until recently, the vast majority of cloud modelling studies have conventionally assumed, implicitly or explicitly, that the soluble material in aerosol particles comprises inorganic components. The emerging complexity of ambient aerosol requires that this description is revisited. It must be remembered that number of activated droplets is dependent on the number distribution of particles of a given type and not directly on the mass loading. Although the acti￾vation of an individual particle is dependent on its (soluble) mass, techniques which coarsely probe component distribu￾tion loadings by mass will not provide adequate insight to predict droplet number. Composition is likely to be impor￾tant only in a limited size range: particles smaller than about 40 nm diameter are unlikely to activate into cloud droplets regardless of their composition and sufficiently large parti￾cles will almost always contain enough soluble material to activate. To predict droplet activation it is necessary to de￾termine size-resolved composition in the 40 to 200 nm size range coupled to information about the mixing-state of the population. No single technique can currently provide all this information. This section reviews the available evidence for the composition of ambient particles from a range of studies and techniques. From combinations of these sources it should eventually be possible to adequately describe the aerosol composition distribution for purposes of CCN and possibly droplet number prediction. Offline analyses of bulk particulate material collected by filter pack and impactor sampling have conventionally been used to determine aerosol mass composition. Applied ana￾lytical techniques can provide information ranging from de￾tailed molecular speciation to aggregate lumped chemical functionality. These techniques have recently been supple￾mented by online instrumentation which may provide addi￾tional information to that available from offline techniques. Some online instruments provide single particle composition information and hence information about which components co-exist in the same particles; their chemical mixing state. Others may provide component mass loadings with high size and time resolution. However, online techniques cannot cur￾rently provide as much detailed speciation information as may be available from offline techniques. 3.1.4.1 Offline bulk measurements showing the complexity of the atmospheric aerosol composition This section presents such features of aerosol composition (organic and inorganic) which may be gained from offline analyses as relate to cloud activation, without attempting an exhaustive review. A summary of some available size segre￾gated chemical compositions is also provided in a form suit￾able for cloud modelling purposes. Despite the wide range of sampling and analytical tech￾niques that have been employed, characterisation of aerosol chemical composition as a function of size is often still in￾complete (Putaud et al., 2004). The chemical composition is diverse, complex and variable with location and condi￾tion. The particles comprise many inorganic and organic compounds ranging in solubility, density and, surface ten￾sion. Thus, comprehensive papers about the aspects of the chemical composition relevant to cloud formation are rare. In order to use Eq. (2), the chemical composition must be “translated” for cloud modelling purposes into concentra￾tion of molecules (organic and inorganic) dissolved in cloud droplets, total insoluble mass and, if present, the concen￾tration of some “critical” species with limited solubility (or slightly soluble species). The degree of dissociation and pH may also be needed (Laaksonen et al., 1998; Lohmann et al., 2004); this is addressed further in Sect. 4. It should be noted that there are exceptions to these requirements when Eq. (1) is used, depending on how water activity is evaluated, see Sect. 4.4. Soluble inorganic components are relatively well un￾derstood; the majority comprising a few inorganic salts (Heintzenberg, 1989), which are relatively well￾characterised in terms of their hygroscopic properties (Clegg et al., 1998; Ansari and Pandis, 2000). The insoluble inor￾ganic fraction can also be important (as in the case of dust aerosol from urban or natural sources) and many different component or element measurements are available. How￾ever, this information is difficult to directly interpret in terms of total insoluble mass. Organic matter has been shown to represent an important fraction of the aerosol mass in different environments, and is routinely measured by means of thermal techniques (Liousse et al., 1996; Jacobson et al., 2000; Putaud et al., 2004). Or￾ganic carbon (OC) and Elemental carbon (EC) (or black car￾bon (BC)) are reported in terms of carbon mass and the trans￾formation to aerosol mass is problematic without knowing the main chemical C structure (Turpin and Lim, 2001; Rus￾www.atmos-chem-phys.net/6/2593/2006/ Atmos. Chem. Phys., 6, 2593–2649, 2006

2602 G. McFiggans et al. aerosol effects on warm cloud activation sell, 2003). The assumption that BC belong to the insoluble Functional group analytical techniques provide an alter fraction of the aerosol has been questioned by recent experi- native approach to traditional individual compound specia- ments showing that thermally refractory fractions of TC can tion methods. These techniques analyse the different types of be efficiently extracted with water (Yu et al., 2004, Mayol- chemical structures(as for example total carboxylic groups, Bracero et al., 2002). Furthermore, OC/BC concentrations total carbonyls, etc. ) but provide little or no information on the individual molecules(Decesari et al., 2000; Maria Experimental studies indicate that, in addition to the et al., 2002). Functional group methods have sometime inorganic components, water-soluble organic compounds been coupled to extraction-classification or separation tech wSOC)in atmospheric aerosol particles are also potentially niques, providing a more comprehensive description of oC important in clouds, and an understanding of organic par- and being able to account for up to 90% of the wSoc titioning in cloud droplets(whether dissolved or present as mass(Decesari et al., 2001; Varga et al., 2001). In partic insoluble inclusions) is crucial to our understanding of their ular, in the functional group analysis approach proposed by possible effects on cloud activation(see for example Fac- Decesari et al. (2000), wsoC is separated into three main chini et al. 1999b Jacobson et al. 2000: Kiss et al.. 200 classes of compounds: neutral compounds Maria et al., 2003). WSOC, as opposed to inorganic aerosol /di-carboxylic acid(MDA)and polycarboxylic acids(Pa) components, comprise hundreds(or even thousands) of in- Quantitative measurements of wsoc by Total Organic Ca dividual species(Saxena and Hildemann, 1996: Maria et al., bon(ToC) analyser and of proton concentration of the 2004, Hamilton et al., 2004, Murphy, 2005; Kanakidou et al., main functional groups contained in each of the three above 2005), each contributing a small fraction of the mass. Sev- mentioned classes by Proton Nuclear Magnetic Resonance eral studies of aerosol WSOC concentration and composition (HNMR)can be used to formulate a set of a few "model have been carried out(Zappoli et al., 1999; Facchini et al., compounds representative of the whole wSoC (Fuzzi et al 1999b, Kiss et al., 2001, 2002; Mayol- Bracero et al., 2002; 2001). A systematic technique for deriving model com Cavalli et al., 2004a, b, Putaud et al., 2004, Sullivan et al., pounds for biomass burning aerosol collected in the Ama- 2004: Xiao and Liu, 2004). Molecular level identification zon has recently been submitted for publication(Decesari and analysis is the traditional goal of aerosol organic analysis et al., 2006). Since the model compounds derived in this (for example IC: Falkovich et al., 2005; IEC-UV: Schkolnik way reproduce quantitatively the average chemical structure et al., 2005; GC-MS: Graham et al., 2002; Pashynska et al., of wSoc it can be argued that they may be used as best 2002; Carvalho et al., 2003; lon et al., 2005), but such indi- guess surrogates in microphysical models involving biomass vidual component approaches only account for a small frac- burning aerosol. Likewise, model mixtures of wSoc for tion of the total aerosol and a long list of compounds present many different types of aerosol in a range of locations are in very small concentration is usually provided. In addition available or their definition is in progress to the analytical procedure, bulk sampling techniques which are frequently employed for such analyses are inappropriate Urban aerosol, Bologna, Italy(Matta et al., 2003; Fuzzi for cloud activation purposes and size-segregated determin etal,2001), tion is necessary( Carvalho et al., 2003; Matta et al., 2003 Dust aerosol, Monte Cimone, Italy(Putaud et al., 2004) Cavalli et al. 2004b: Putaud et al.. 2004: Falkovich et al 2005) Clean marine aerosol, Mace Head, Ireland(Cavalli The representation of aerosol composition therefore et al., 2004b) presents a dilemma; it is evident that the aerosol wSoc Biomass burning aerosol, Rondonia, Brazil (Decesari cannot be correctly represented by molecules accounting fo etal,2006), only a small fraction of the total carbon mass, but a repre- sentation of participating species is required for a fundamen ACE Asia, Chinese outflow, Gosan, Jeju Island, Korea tal prediction of cloud activation. Frequently, due to the Topping et al., 2004), complexity, the wSoC chemical composition is reduced fo modelling purposes to one or two"representative"species or Boreal forest aerosol, Hyytiala, Finland( Cavalli et al surrogate molecules selected from the long list of compounds 2004a, Decesari et al., 2006) Ity of wSOC and the wide range of physical properties rele- tion concerning both inw studies have provided informa- and organic aerosol chemi vant to activation, an arbitrary choice of representative com- cal composition which can be directly used by cloud mod- pounds can fail in reproducing relevant physical and chemi- els. These papers provide a comprehensive description of the al properties. For the above reasons, a"realistic"represen- chemical composition of different aerosol types as a function tation of wsoc is necessary for cloud modelling purposes, of size( Chan et al., 1999; Zappoli et al., 1999, Pakkanen but it is difficult to achieve through any individual analyti- et al., 2001; Putaud et al., 2000; Temesi et al., 2001; Maria cal methodology or by choosing surrogate chemical compo- et al., 2003; Sellegri et al., 2003; Cabada et al., 2004, Chic sitions from a list of compounds detected in the aerosols et al., 2004; Sardar et al., 2005) tmos.Chem.Phvs.6.2593-26492006 www.atmos-chem-phys.net/6/2593/2006/

2602 G. McFiggans et al.: Aerosol effects on warm cloud activation sell, 2003). The assumption that BC belong to the insoluble fraction of the aerosol has been questioned by recent experi￾ments showing that thermally refractory fractions of TC can be efficiently extracted with water (Yu et al., 2004; Mayol￾Bracero et al., 2002). Furthermore, OC/BC concentrations are strongly size dependent. Experimental studies indicate that, in addition to the inorganic components, water-soluble organic compounds (WSOC) in atmospheric aerosol particles are also potentially important in clouds, and an understanding of organic par￾titioning in cloud droplets (whether dissolved or present as insoluble inclusions) is crucial to our understanding of their possible effects on cloud activation (see for example Fac￾chini et al., 1999b; Jacobson et al., 2000; Kiss et al., 2001; Maria et al., 2003). WSOC, as opposed to inorganic aerosol components, comprise hundreds (or even thousands) of in￾dividual species (Saxena and Hildemann, 1996; Maria et al., 2004; Hamilton et al., 2004; Murphy, 2005; Kanakidou et al., 2005), each contributing a small fraction of the mass. Sev￾eral studies of aerosol WSOC concentration and composition have been carried out (Zappoli et al., 1999; Facchini et al., 1999b; Kiss et al., 2001, 2002; Mayol-Bracero et al., 2002; Cavalli et al., 2004a,b; Putaud et al., 2004; Sullivan et al., 2004; Xiao and Liu, 2004). Molecular level identification and analysis is the traditional goal of aerosol organic analysis (for example IC: Falkovich et al., 2005; IEC-UV: Schkolnik et al., 2005; GC-MS: Graham et al., 2002; Pashynska et al., 2002; Carvalho et al., 2003; Ion et al., 2005), but such indi￾vidual component approaches only account for a small frac￾tion of the total aerosol and a long list of compounds present in very small concentration is usually provided. In addition to the analytical procedure, bulk sampling techniques which are frequently employed for such analyses are inappropriate for cloud activation purposes and size-segregated determina￾tion is necessary (Carvalho et al., 2003; Matta et al., 2003; Cavalli et al., 2004b; Putaud et al., 2004; Falkovich et al., 2005). The representation of aerosol composition therefore presents a dilemma; it is evident that the aerosol WSOC cannot be correctly represented by molecules accounting for only a small fraction of the total carbon mass, but a repre￾sentation of participating species is required for a fundamen￾tal prediction of cloud activation. Frequently, due to the its complexity, the WSOC chemical composition is reduced for modelling purposes to one or two “representative” species or surrogate molecules selected from the long list of compounds detected in the atmosphere. However, due to the complex￾ity of WSOC and the wide range of physical properties rele￾vant to activation, an arbitrary choice of representative com￾pounds can fail in reproducing relevant physical and chemi￾cal properties. For the above reasons, a “realistic” represen￾tation of WSOC is necessary for cloud modelling purposes, but it is difficult to achieve through any individual analyti￾cal methodology or by choosing surrogate chemical compo￾sitions from a list of compounds detected in the aerosols. Functional group analytical techniques provide an alter￾native approach to traditional individual compound specia￾tion methods. These techniques analyse the different types of chemical structures (as for example total carboxylic groups, total carbonyls, etc.), but provide little or no information on the individual molecules (Decesari et al., 2000; Maria et al., 2002). Functional group methods have sometime been coupled to extraction-classification or separation tech￾niques, providing a more comprehensive description of OC and being able to account for up to 90% of the WSOC mass (Decesari et al., 2001; Varga et al., 2001). In partic￾ular, in the functional group analysis approach proposed by Decesari et al. (2000), WSOC is separated into three main classes of compounds: neutral compounds (NC), mono- /di-carboxylic acid (MDA) and polycarboxylic acids (PA). Quantitative measurements of WSOC by Total Organic Car￾bon (TOC) analyser and of proton concentration of the main functional groups contained in each of the three above mentioned classes by Proton Nuclear Magnetic Resonance (HNMR) can be used to formulate a set of a few “model” compounds representative of the whole WSOC (Fuzzi et al., 2001). A systematic technique for deriving model com￾pounds for biomass burning aerosol collected in the Ama￾zon has recently been submitted for publication (Decesari et al., 2006). Since the model compounds derived in this way reproduce quantitatively the average chemical structure of WSOC it can be argued that they may be used as best￾guess surrogates in microphysical models involving biomass burning aerosol. Likewise, model mixtures of WSOC for many different types of aerosol in a range of locations are available or their definition is in progress: . Urban aerosol, Bologna, Italy (Matta et al., 2003; Fuzzi et al., 2001), . Dust aerosol, Monte Cimone, Italy (Putaud et al., 2004), . Clean marine aerosol, Mace Head, Ireland (Cavalli et al., 2004b), . Biomass burning aerosol, Rondonia, Brazil (Decesari et al., 2006), . ACE Asia, Chinese outflow, Gosan, Jeju Island, Korea (Topping et al., 2004), . Boreal forest aerosol, Hyytial¨ a, Finland ( ¨ Cavalli et al., 2004a; Decesari et al., 2006). In summary, only a few studies have provided informa￾tion concerning both inorganic and organic aerosol chemi￾cal composition which can be directly used by cloud mod￾els. These papers provide a comprehensive description of the chemical composition of different aerosol types as a function of size (Chan et al., 1999; Zappoli et al., 1999; Pakkanen et al., 2001; Putaud et al., 2000; Temesi et al., 2001; Maria et al., 2003; Sellegri et al., 2003; Cabada et al., 2004; Chio et al., 2004; Sardar et al., 2005). Atmos. Chem. Phys., 6, 2593–2649, 2006 www.atmos-chem-phys.net/6/2593/2006/