2608 G. McFiggans et al. aerosol effects on warm cloud activation Table 3. Hygroscopic behaviour of Aitken mode particles in various environments upper Aitken mode size range(50-80 nm) dominant mode second mode G Fp(90%) fraction G Fp(90%) fraction References occurrence occurrence McMurry and Stolzenburg (1989) Cocker et al. (2001) 1.15-1.43 pto11-90%1.00-1.12 10-89% Zhang et a.(1993) 100% Baltensperger et al. (2002) Svenningsson et al. (1992) Pitchford and McMurry(1994) son et al. (1994) Continental1.32-1.5390-100%28-97%1.05-1.1510-90%3-72% Svenningsson et al.(1997) Polluted Swietlicki et al. (1999, 2000) Hameri et al. (2001) Busch et al. (2002) Boy et al. (2003) Ferron et al. (2005) 1.40-1.5 100% 75-90% bimodal growth(GFD14-1.6) Swietlicki et al. (2000) observed at Izana, Tenerife, 2357m Weingartner et al. (2002) a.s. l. in 47% of clean cases .15-1.20 >97%87-95%~1.07 15-35% dominated Rissler et al. (2004) Remote 1.42-1.75 100% >75%205-2.13 7% <7% Berg et al. (1998b) Marine Swietlicki et al. (2000) Zhou et al. (2001) G FD values of individual aerosol compounds are sum- as 1.65 at 90% RH for highly efficient grate burners leav hygroscopic particles include fresh mineral dust, fresh diesel elemental carbon are expected to have a smale a enor marised in Table 2. Inorganic salts are highly hygroscopic ing only inorganic residue. However, particles from open or with NaCl at the upper end of the growth scale. Non- smouldering flames with significant amounts of organic and engine exhaust(soot- mainly elemental carbon and water insoluble organic species, WINSOL )and fresh petrol en- gine exhaust(WINSOL-dominated). (Note, that Petzold e An overview of growth modes of atmospheric aerosols ob- rved in various environments is given in Table 3 for the up- et al.(2005)recently reported the strong dependence of G FD per Aitken mode size range(50-80 nm)and Table 4 for the and CCN activation of gas turbine combustor particles on lower accumulation mode size range(100-150 nm). These combustion conditions and resultant content of volatile and size ranges were chosen to cover the critical size range for involatile organic compounds and sulphuric acid). Character- CCN activation in clouds. The trend is for 50 nm particles to isation of the water-soluble organic aerosol fraction is more be less hygroscopic than particles at 150 or 165 nm. For the complicated. Pure organic acids range from non-hygroscopic studies listed in the table the growth factor at about 150nm to GFD values similar to inorganic salts. However, G FD averages about 0. I larger than at 50 nm. This may be be of -1. 15 at 90% RH, as observed for secondary organic cause of greater organic content near 50 nm(see section on aerosols formed in reaction chamber experiments and for bulk composition). Whatever the reason, less water uptake humic-like substances, are probably more representative of at smaller sizes implies slightly less sensitivity of activation organic mixtures in aged atmospheric particles, except for to updraught velocity, since a faster updraught that activates urban aerosols, where WINSOL are likely to be dominant. more particles will encounter particles that are less hygro- GFD values of biomass burning particles may be as high scopic. For example, if an air parcel that would activate 120 nm particles in a given updraught instead encounters an tmos.Chem.Phvs.6.2593-26492006 www.atmos-chem-phys.net/6/2593/2006/
2608 G. McFiggans et al.: Aerosol effects on warm cloud activation Table 3. Hygroscopic behaviour of Aitken mode particles in various environments. upper Aitken mode size range (∼50–80 nm) dominant mode second mode frequency frequency GFD(90%) of fraction GFD(90%) of fraction References occurrence occurrence McMurry and Stolzenburg (1989) Cocker et al. (2001) Urban 1.15–1.43 up to 11–90% 1.00–1.12 up to 10–89% Zhang et al. (1993) 100% 100% Baltensperger et al. (2002) Chen et al. (2003) Ferron et al. (2005) Svenningsson et al. (1992) Zhang et al. (1993) Pitchford and McMurry (1994) Svenningsson et al. (1994) Continental 1.32–1.53 90–100% 28–97% 1.05–1.15 10–90% 3–72% Svenningsson et al. (1997) Polluted Swietlicki et al. (1999, 2000) Hameri et al. ¨ (2001) Busch et al. (2002) Boy et al. (2003) Ferron et al. (2005) Free 1.40–1.55 100% 75–90% bimodal growth (GFD∼1.4–1.6) Swietlicki et al. (2000) Troposphere observed at Izana, Tenerife, 2357 m ˜ Weingartner et al. (2002) a.s.l. in 47% of clean cases Biogenically 1.15–1.20 >97% 87–95% ∼1.07 15–35% <12% Zhou et al. (2002) dominated Rissler et al. (2004) Remote 1.42–1.75 100% >75% 2.05-2.13 3–7% <7% Berg et al. (1998b) Marine Swietlicki et al. (2000) Zhou et al. (2001) GFD values of individual aerosol compounds are summarised in Table 2. Inorganic salts are highly hygroscopic with NaCl at the upper end of the growth scale. Nonhygroscopic particles include fresh mineral dust, fresh diesel engine exhaust (soot – mainly elemental carbon and waterinsoluble organic species, “WINSOL”) and fresh petrol engine exhaust (WINSOL-dominated). (Note, that Petzold et al. (2005) recently reported the strong dependence of GFD and CCN activation of gas turbine combustor particles on combustion conditions and resultant content of volatile and involatile organic compounds and sulphuric acid). Characterisation of the water-soluble organic aerosol fraction is more complicated. Pure organic acids range from non-hygroscopic to GFD values similar to inorganic salts. However, GFD of ∼1.15 at 90% RH, as observed for secondary organic aerosols formed in reaction chamber experiments and for humic-like substances, are probably more representative of organic mixtures in aged atmospheric particles, except for urban aerosols, where WINSOL are likely to be dominant. GFD values of biomass burning particles may be as high as 1.65 at 90% RH for highly efficient grate burners leaving only inorganic residue. However, particles from open or smouldering flames with significant amounts of organic and elemental carbon are expected to have a smaller GFD. An overview of growth modes of atmospheric aerosols observed in various environments is given in Table 3 for the upper Aitken mode size range (∼50–80 nm) and Table 4 for the lower accumulation mode size range (∼100–150 nm). These size ranges were chosen to cover the critical size range for CCN activation in clouds. The trend is for 50 nm particles to be less hygroscopic than particles at 150 or 165 nm. For the studies listed in the table the growth factor at about 150 nm averages about 0.1 larger than at 50 nm. This may be because of greater organic content near 50 nm (see section on bulk composition). Whatever the reason, less water uptake at smaller sizes implies slightly less sensitivity of activation to updraught velocity, since a faster updraught that activates more particles will encounter particles that are less hygroscopic. For example, if an air parcel that would activate 120 nm particles in a given updraught instead encounters an 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 Table 4. Hygroscopic behaviour of accumulation mode particles in various environments lower accumulation mode size range(100-150nm) dominant mode second mode GFD(90%) fraction G FD(90%) fraction References occurrence occurrence McMurry and Stolzenburg (1989) Cocker et al. (2001) Urban 123-1.50 90%1.00-1.14 up to 10-84% Zhang et al. (1993) Ferron et al. (2005) Svenningsson et al. (1992 Covert and Heinzenberg(1993) Zhang et al. (1993) itchford and McMurry(1994) Svenningsson et al. (1994) Continental 1. 41-1.6 100% 37-100%1.03-1.18 2-63% Svenningsson et al. (1997) Swietlicki et al. (1999, 2000) Hameri et al. (2001) Busch et al. (2002) Boy et al.(2003) Ferron et al. (2005) ree <1.3 M15% Weingartner(2002) Biogenically 1. 20-1. 25 90-96% Zhou et al. (2002) dominated Rissler et al. (2004) Remote 147-1.78 1009 >80%2.06-2.14 13-40% 15% Berg et al. (1998b) Marine Swietlicki et al. (2000) Zhou et al. (2001) updraught which would activate particles of 75 nm of the is often similar but varies strongly from location to location same composition, the 75 nm particles would not activate and as a function of time. G Fp of the intermediately hy since they are not of the same composition but are less hy- groscopic background particles tend to be somewhat smaller groscopic. Hence an even greater updraught would be re- than the dominant mode in continental polluted environ- quired to activate the 75 nm particles. The compilations in ments, indicating the hygroscopicity of the background par- Tables 3 and 4 are rather a reflection of ranges of observed ticles in urban areas is reduced by the presence of less hygro- values than a full climatology. The tables contain the mean scopic compounds growth factor, the frequency of occurrence and the fraction of total particles observed for the dominant and the second Continental polluted aerosols have been investigated in most important growth mode. Reduction of ambient growth many locations ranging from near-urban to remote and hence distributions to two growth modes may be an oversimplifica- the corresponding growth characteristics provided in Ta- tion in some cases, but many studies report only two modes, bles 3 and 4 cover a wide range of air masses. The read- and the actual growth distribution may often be reasonably ily hygroscopic background mode generally dominates con- well represented by two growth modes in conjunction with tinental polluted aerosols, though considerable fractions of corresponding growth spread factors marginally hygroscopic particles may be found in pro imity to urban areas. The background mode is always rban aerosols are often characterised by an external mix- present, whereas marginally hygroscopic particles are not ture(see Sect. 3.2.3)of distinct modes of intermediately hy- always found. There is some trend of decreasing fraction groscopic background particles and marginally hygroscopic of marginally hygroscopic particles with increasing distance particles from local emissions. The abundance of both modes from major anthropogenic sources, but they may still 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 2609 Table 4. Hygroscopic behaviour of accumulation mode particles in various environments. lower accumulation mode size range (∼100–150 nm) dominant mode second mode frequency frequency GFD(90%) of fraction GFD(90%) of fraction References occurrence occurrence McMurry and Stolzenburg (1989) Cocker et al. (2001) Urban 1.23–1.50 100% 16–90% 1.00–1.14 up to 10–84% Zhang et al. (1993) 100% Baltensperger et al. (2002) Chen et al. (2003) Santarpia et al. (2004) Ferron et al. (2005) Svenningsson et al. (1992) Covert and Heintzenberg (1993) Zhang et al. (1993) Pitchford and McMurry (1994) Svenningsson et al. (1994) Continental 1.41–1.64 100% 37–100% 1.03–1.18 10–90% 2–63% Svenningsson et al. (1997) Polluted Swietlicki et al. (1999, 2000) Hameri et al. ¨ (2001) Busch et al. (2002) Boy et al. (2003) Ferron et al. (2005) Free 1.62 100% ∼85% <1.3 – ∼15% Weingartner (2002) Troposphere Biogenically 1.20–1.25 >93% 90–96% ∼1.08 ∼25% <7% Zhou et al. (2002) dominated Rissler et al. (2004) Remote 1.47–1.78 100% >80% 2.06–2.14 13–40% <15% Berg et al. (1998b) Marine Swietlicki et al. (2000) Zhou et al. (2001) updraught which would activate particles of 75 nm of the same composition, the 75 nm particles would not activate since they are not of the same composition but are less hygroscopic. Hence an even greater updraught would be required to activate the 75 nm particles. The compilations in Tables 3 and 4 are rather a reflection of ranges of observed values than a full climatology. The tables contain the mean growth factor, the frequency of occurrence and the fraction of total particles observed for the dominant and the second most important growth mode. Reduction of ambient growth distributions to two growth modes may be an oversimplification in some cases, but many studies report only two modes, and the actual growth distribution may often be reasonably well represented by two growth modes in conjunction with corresponding growth spread factors. Urban aerosols are often characterised by an external mixture (see Sect. 3.2.3) of distinct modes of intermediately hygroscopic background particles and marginally hygroscopic particles from local emissions. The abundance of both modes is often similar but varies strongly from location to location and as a function of time. GFD of the intermediately hygroscopic background particles tend to be somewhat smaller than the dominant mode in continental polluted environments, indicating the hygroscopicity of the background particles in urban areas is reduced by the presence of less hygroscopic compounds. Continental polluted aerosols have been investigated in many locations ranging from near-urban to remote and hence the corresponding growth characteristics provided in Tables 3 and 4 cover a wide range of air masses. The readily hygroscopic background mode generally dominates continental polluted aerosols, though considerable fractions of marginally hygroscopic particles may be found in proximity to urban areas. The background mode is always present, whereas marginally hygroscopic particles are not always found. There is some trend of decreasing fraction of marginally hygroscopic particles with increasing distance from major anthropogenic sources, but they may still be www.atmos-chem-phys.net/6/2593/2006/ Atmos. Chem. Phys., 6, 2593–2649, 2006
2610 G. McFiggans et al. aerosol effects on warm cloud activation found after 5000 km or 5 days of transport(Heintzenberg and Hygroscopic properties of remote marine aerosols have Covert, 1987). The conditions and processes favouring the been investigated in the Pacific, Southern, north-eastern At- removal or transformation of marginally hygroscopic parti- lantic and central Arctic Oceans. Values given in Tables 3 cles are still to be elucidated and 4 exclude episodes of significant anthropogenic influ Only a few of the HTDMA measurements of free tro- ence. The aerosol in remote environment shows pospheric aerosols unaffected by recent anthropogenic in- mostly unimodal growth with growth factors larger than fluence taken as part of the Global Atmosphere Watch in any other location. Observed growth factors are of- (GAW) programme at the Izana(Tenerife, 2367m a.s.l. )and ten between the growth factors of pure sulphate and pure Jungfraujoch( Switzerland, 3580 m a.s. 1 ) stations have been sodium chloride particles(cp. Table 2), indicating internal or published. Values shown in Tables 3 and 4 include only pe- quasi-internal mixture of sulphate, sea salt and organics(see riods where the stations were unaffected by the mixed plan- Sect. 3.2.3). Externally mixed sea salt dominated particl etary boundary layer, and they exclude Sahara dust events with growth factors larger than two are occasionally observed connected with externally mixed mineral dust particles and are linked with high wind speeds on a regional scale with occasionally observed at the Jungfraujoch station( Weingart- a critical value of the order of 10 m/s. They are more frequent ner et al., 2001). The free tropospheric aerosol is dominated at larger sizes but they are also found in the Aitken mode size by readily hygroscopic particles. Bimodal growth distribu- range. Events of anthropogenic pollution were linked with tions with G FD-1.4 and -1. 6 were frequently observed decreasing growth factors of the dominant mode and with for Aitken mode particles at the Izana station, whereas nar- occurrence of an externally mixed marginally or moderately row growth distributions were characteristic for the Jungfrau- hygroscopic mode. Occasionally an externally mixed mode joch station. The trend of increasing growth factors from of moderately hygroscopic particles was reported in the ab- urban through continental polluted to free tropospheric air sence of anthropogenic influence in the pack ice covered cen- masses indicates that, with the progression of atmospheric tral Arctic Ocean during summer(Zhou et al., 2001) but the ageing processes, readily hygroscopic inorganic salts dom- origin or composition of these particles remained unclear inate moderately hygroscopic organic compounds. This is consistent with the observation that accumulation mode par- 3.2.2 CCN measurement ticles are generally more hygroscopic than Aitken mode par ticles, indicating that the increased hygroscopicity of accu- Measurement of the cloud condensation nuclei, or CCN, mulation mode particles is mainly due to their increased frac- provides a direct estimation of the propensity of aerosol to tion of inorganic salts. Such dominance of inorganic com- form cloud droplets. This propensity is reported as a poten- pounds and low contribution of even highly oxygenated or- tial cloud droplet concentration, realized at a specified wa- ganic components to equilibrium water content(and hence ter vapour supersaturation, or as a supersaturation-dependent G FD) in aged aerosol was also reported by McFiggans et al. spectrum of potential droplet concentrations. The latter is (2005) commonly referred to as the CCn activation spectrum. with Biogenically dominated aerosols were investigated in the the exception of the technique developed by Hudson(1989), pristine Amazonian rain forest. Corresponding values re- CCN activation spectra are obtained from step-wise scans of ported in Tables 3 and 4 exclude episodes of clear influence the imposed supersaturation and presented as cumulative dis- from biomass burning. These aerosols are dominated by tributions. Activation spectra can also be derived from mea- ganic compounds and hence the growth factors are moderate. surements of an aerosol size spectrum and composition, but A minor fraction of marginally hygroscopic particles is oc- a theoretical relationship between dry size and critical super- casionally present. Air masses influenced by recent or aged saturation(e.g. Fig. 2)is necessary for such calculations. Ac- biomass burning showed increased hygroscopic growth fac- cordingly, CCN activation spectra(measured or derived)are tors of the dominant mode or presence of externally mixed classifiable as distribution-dependent properties(Sect. 3. 2) particles with larger growth factors, but a larger fraction of Comparison of a measured activation spectrum with a de- marginally hygroscopic particles was also observed. This rived spectrum constitutes a validation of the latter, provided indicates that efficient biomass burning releases readily hy- that uncertainties in the measurements, as well as error intro- groscopic inorganic residue, but that also marginally hygro- duced by simplifying assumptions(e.g. Kohler theory, mix- scopic particles are produced during the smouldering phase. ing state, particle composition, etc. ) are quantified. Such ex- Impacts of biomass burning have been reported elsewhere ercises establish reliable data sets for cloud model initialise- showing both trends of increasing( Cocker et al., 2001)and tion(e.g. Brenguier, 2003), and can also reveal limitations of decreasing hygroscopicity(Carrico et al., 2005). Measure- the measurements( Chuang et al., 2000; Snider et al., 2003) ments in other locations with significant biogenic emission Over the past four decades several different CCN instru- also show moderate hygroscopic growth factors associated ments have emerged, each with advantages and drawbacks with large organic aerosol fractions(Hameri et al, 2001; inherent to their design. These designs are the isothermal Aklilu and Mozurkewich, 2004; Carrico et al., 2005) haze chamber (Laktionov, 1972), the diffusion tube(Leaitch and Megaw, 1982), the CCn"remover(Ji et al., 1998 tmos.Chem.Phvs.6.2593-26492006 www.atmos-chem-phys.net/6/2593/2006/
2610 G. McFiggans et al.: Aerosol effects on warm cloud activation found after 5000 km or 5 days of transport (Heintzenberg and Covert, 1987). The conditions and processes favouring the removal or transformation of marginally hygroscopic particles are still to be elucidated. Only a few of the HTDMA measurements of free tropospheric aerosols unaffected by recent anthropogenic in- fluence taken as part of the Global Atmosphere Watch (GAW) programme at the Izana (Tenerife, 2367 m a.s.l.) and ˜ Jungfraujoch (Switzerland, 3580 m a.s.l.) stations have been published. Values shown in Tables 3 and 4 include only periods where the stations were unaffected by the mixed planetary boundary layer, and they exclude Sahara dust events connected with externally mixed mineral dust particles as occasionally observed at the Jungfraujoch station (Weingartner et al., 2001). The free tropospheric aerosol is dominated by readily hygroscopic particles. Bimodal growth distributions with GFD∼1.4 and ∼1.6 were frequently observed for Aitken mode particles at the Izana station, whereas nar- ˜ row growth distributions were characteristic for the Jungfraujoch station. The trend of increasing growth factors from urban through continental polluted to free tropospheric air masses indicates that, with the progression of atmospheric ageing processes, readily hygroscopic inorganic salts dominate moderately hygroscopic organic compounds. This is consistent with the observation that accumulation mode particles are generally more hygroscopic than Aitken mode particles, indicating that the increased hygroscopicity of accumulation mode particles is mainly due to their increased fraction of inorganic salts. Such dominance of inorganic compounds and low contribution of even highly oxygenated organic components to equilibrium water content (and hence GFD) in aged aerosol was also reported by McFiggans et al. (2005). Biogenically dominated aerosols were investigated in the pristine Amazonian rain forest. Corresponding values reported in Tables 3 and 4 exclude episodes of clear influence from biomass burning. These aerosols are dominated by organic compounds and hence the growth factors are moderate. A minor fraction of marginally hygroscopic particles is occasionally present. Air masses influenced by recent or aged biomass burning showed increased hygroscopic growth factors of the dominant mode or presence of externally mixed particles with larger growth factors, but a larger fraction of marginally hygroscopic particles was also observed. This indicates that efficient biomass burning releases readily hygroscopic inorganic residue, but that also marginally hygroscopic particles are produced during the smouldering phase. Impacts of biomass burning have been reported elsewhere showing both trends of increasing (Cocker et al., 2001) and decreasing hygroscopicity (Carrico et al., 2005). Measurements in other locations with significant biogenic emission also show moderate hygroscopic growth factors associated with large organic aerosol fractions (Hameri et al. ¨ , 2001; Aklilu and Mozurkewich, 2004; Carrico et al., 2005). Hygroscopic properties of remote marine aerosols have been investigated in the Pacific, Southern, north-eastern Atlantic and central Arctic Oceans. Values given in Tables 3 and 4 exclude episodes of significant anthropogenic influence. The aerosol in remote marine environment shows mostly unimodal growth with growth factors larger than in any other location. Observed growth factors are often between the growth factors of pure sulphate and pure sodium chloride particles (cp. Table 2), indicating internal or quasi-internal mixture of sulphate, sea salt and organics (see Sect. 3.2.3). Externally mixed sea salt dominated particles with growth factors larger than two are occasionally observed and are linked with high wind speeds on a regional scale with a critical value of the order of 10 m/s. They are more frequent at larger sizes but they are also found in the Aitken mode size range. Events of anthropogenic pollution were linked with decreasing growth factors of the dominant mode and with occurrence of an externally mixed marginally or moderately hygroscopic mode. Occasionally an externally mixed mode of moderately hygroscopic particles was reported in the absence of anthropogenic influence in the pack ice covered central Arctic Ocean during summer (Zhou et al., 2001) but the origin or composition of these particles remained unclear. 3.2.2 CCN measurement Measurement of the cloud condensation nuclei, or CCN, provides a direct estimation of the propensity of aerosol to form cloud droplets. This propensity is reported as a potential cloud droplet concentration, realized at a specified water vapour supersaturation, or as a supersaturation-dependent spectrum of potential droplet concentrations. The latter is commonly referred to as the CCN activation spectrum. With the exception of the technique developed by Hudson (1989), CCN activation spectra are obtained from step-wise scans of the imposed supersaturation and presented as cumulative distributions. Activation spectra can also be derived from measurements of an aerosol size spectrum and composition, but a theoretical relationship between dry size and critical supersaturation (e.g. Fig. 2) is necessary for such calculations. Accordingly, CCN activation spectra (measured or derived) are classifiable as distribution-dependent properties (Sect. 3.2). Comparison of a measured activation spectrum with a derived spectrum constitutes a validation of the latter, provided that uncertainties in the measurements, as well as error introduced by simplifying assumptions (e.g. Kohler theory, mix- ¨ ing state, particle composition, etc.), are quantified. Such exercises establish reliable data sets for cloud model initialisation (e.g. Brenguier, 2003), and can also reveal limitations of the measurements (Chuang et al., 2000; Snider et al., 2003). Over the past four decades several different CCN instruments have emerged, each with advantages and drawbacks inherent to their design. These designs are the isothermal haze chamber (Laktionov, 1972), the diffusion tube (Leaitch and Megaw, 1982), the CCN “remover” (Ji et al., 1998), 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 2611 the continuous-fow chamber(Sinnarwalla and Alofs, 1972: (cf. Rogers and Yau, 1989), the duration of the peak supersat Chuang et al., 2000, Otto et al., 2002, Hudson, 1989, Van- uration varies with vertical velocity and aerosol background. Reken et al., 2003; Roberts and Nenes, 2005), the chemical A representative value is 10 s and is consistent with growth diffusion chamber(Twomey, 1959), and the static thermal times in many CCN instruments. Although speculation does diffusion chamber (Squires and Twomey, 1966, Wieland, exist that a mismatch of the instrument-imposed and natural 56: Lala and Jiusto. 1977: Oliveira and vali. 1995: Giebl times can lead to bias in the estimation of ccn activation et al., 2002; Snider et al., 2003). Nenes et al.(2001a)has( Chuang et al. 1997), this issue has not been addressed ex- evaluated the performance of some of these instruments with perimentally. Investigations planned for a large continuous- a computational model. With the exception of the technique flow instrument( Stratmann et al., 2004)may contribute to developed by(Hudson, 1989), CCN activation spectra are resolution of this issue obtained from step-wise scans of the imposed supersatura tion and presented as cumulative distributions. In the for- 3.2.3 Aerosol mixing-state with respect to water uptake mer technique the size distribution of activated droplets measured and used to derive the differential CCN activation A complicated composition mixing state is expected in the ambient atmosphere due to the various primary particle Since the ccn data are couched in terms of the maxi- sources and numerous processes altering particle compo- mum supersaturation experienced by particles within a CCn sition. Each kind of internally mixed particle may con- instrument (i.e. within the particle growth chamber), re- tain a range of fractional contributions of each constituent, cent emphasis has been placed on accurate specification of hereinafter referred to as quasi-internal mixtures. External this operationally-defined supersaturation. Seminal work on mixing of differently hygroscopic compounds results in dis- this problem was ambiguous showing either that the max- tinct growth modes, whereas quasi-internal mixing results in imum supersaturation derived using chamber temperatures spread of observed growth factors around the growth fac- exceeded(Katz and Kocmond, 1973)or that it was consistent tor corresponding to the mean composition. This descrip- with the maximum supersaturation derived from a particle tion is consistent with the discussion of composition pre- size sufficient for activation( Gerber et al. 1977; Alofs et al., sented in Sect. 3. 1. 4.2.1. Detection of different G FD modes 1979). Consistency in this context implies that measured or a continuous G FD distribution with a HTDMA implies temperatures, transformed to a maximum chamber supersat- that the aerosol is to some extent externally mixed. How uration via a model of the temperature and vapour fields, is ever, detection of a single growth factor does not exclude consistent with the chamber supersaturation derived by in- the presence of externally mixed particles, as different mix- putting the minimum dry size necessary for activation into tures may still have similar growth factors. At least two dis- Kohler theory. The laboratory work of Leaitch et al.(1999): tinct growth factor modes are frequently observed in various Snider et al. (2003); Bilde et al. (2003); Bilde and Svennings- environments, showing that extemal mixing is widespread son(2004)reveal relationships of the form Seff=k Snom with Some studies even report three to six simultaneous growth Snom and Seff defined in terms of chamber temperature and modes Berg et al., 1998a; Cocker et al., 2001; Carrico et al particle size, respectively, and K varying between 0.8 and 2005). In both Aitken and accumulation mode size ranges, 5. This correction is specific to measurements made with an externally mixed mode of marginally hygroscopic parti- the University of Wyoming CCN instrument and is compara- cles (G FDil. 15 at 90%RH is frequently found in most en- ble to that made for other static diffusion instruments( rissler vironments. Numerous studies combining different methods et al., 2004). Since determination of Snom requires measure- indicate that these marginally hygroscopic particles are dom- ment and control of temperatures with an accuracy of better inated by elemental carbon( Covert and Heintzenberg, 1984, than 0. 1C, especially at supersaturations less than 0.2%, the Heintzenberg and Covert, 1987; Zhang et al., 1993, MaBling documented range of K is partially a reflection of the diffi- et al., 2003) and/or mineral dust(Zhang et al., 1993; Dick The ng to the temperature measurement/control et al., 1998, 2000; Weingartner et al., 2001 ) as can be ex- propagating from the measurement of test particle size used (Table 2). Mineral dust appears to be found in particles with in the determination of Seff small growth factors only, whereas elemental carbon is also The duration of particle exposure to the effective maxi- found in particles with larger growth factors. Small amounts mum supersaturation is also a parameter which deserves con- of inorganic salts or organic compounds may also be present especially important if kinetic control of water uptake is am- ticles are nearly ubiquitous, the dominating growth mode in plified by particles with coatings that produce low or var- most environments is readily hygroscopic with G FD values led water accommodation coefficients( see Sect. 4. 1). Ideally ranging from about 1. 2 to 1.8 at 90% RH. Inorganic salts his time interval should mimic that which occurs during acti- and organic compounds are understood to be the main ingre- ation in an updraft. According to parcel model calculations dients of these, quasi-internally mixed, readily hygroscopic 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 2611 the continuous-flow chamber (Sinnarwalla and Alofs, 1972; Chuang et al., 2000; Otto et al., 2002; Hudson, 1989; VanReken et al., 2003; Roberts and Nenes, 2005), the chemical diffusion chamber (Twomey, 1959), and the static thermal diffusion chamber (Squires and Twomey, 1966; Wieland, 1956; Lala and Jiusto, 1977; Oliveira and Vali, 1995; Giebl et al., 2002; Snider et al., 2003). Nenes et al. (2001a) has evaluated the performance of some of these instruments with a computational model. With the exception of the technique developed by (Hudson, 1989), CCN activation spectra are obtained from step-wise scans of the imposed supersaturation and presented as cumulative distributions. In the former technique the size distribution of activated droplets is measured and used to derive the differential CCN activation spectrum. Since the CCN data are couched in terms of the maximum supersaturation experienced by particles within a CCN instrument (i.e. within the particle growth chamber), recent emphasis has been placed on accurate specification of this operationally-defined supersaturation. Seminal work on this problem was ambiguous showing either that the maximum supersaturation derived using chamber temperatures exceeded (Katz and Kocmond, 1973) or that it was consistent with the maximum supersaturation derived from a particle size sufficient for activation (Gerber et al., 1977; Alofs et al., 1979). Consistency in this context implies that measured temperatures, transformed to a maximum chamber supersaturation via a model of the temperature and vapour fields, is consistent with the chamber supersaturation derived by inputting the minimum dry size necessary for activation into Kohler theory. The laboratory work of ¨ Leaitch et al. (1999); Snider et al. (2003); Bilde et al. (2003); Bilde and Svenningsson (2004) reveal relationships of the form Seff=κSnom with Snom and Seff defined in terms of chamber temperature and particle size, respectively, and κ varying between 0.8 and 0.5. This correction is specific to measurements made with the University of Wyoming CCN instrument and is comparable to that made for other static diffusion instruments (Rissler et al., 2004). Since determination of Snom requires measurement and control of temperatures with an accuracy of better than 0.1◦C, especially at supersaturations less than 0.2%, the documented range of κ is partially a reflection of the diffi- culty of conforming to the temperature measurement/control strictures. The value of κ is also variable because of error propagating from the measurement of test particle size used in the determination of Seff. The duration of particle exposure to the effective maximum supersaturation is also a parameter which deserves consideration in an evaluation of CCN data, or in their use in closure studies. Residence time in the CCN instrument is especially important if kinetic control of water uptake is amplified by particles with coatings that produce low or varied water accommodation coefficients (see Sect. 4.1). Ideally this time interval should mimic that which occurs during activation in an updraft. According to parcel model calculations (cf. Rogers and Yau, 1989), the duration of the peak supersaturation varies with vertical velocity and aerosol background. A representative value is ∼10 s and is consistent with growth times in many CCN instruments. Although speculation does exist that a mismatch of the instrument-imposed and natural times can lead to bias in the estimation of CCN activation (Chuang et al., 1997), this issue has not been addressed experimentally. Investigations planned for a large continuous- flow instrument (Stratmann et al., 2004) may contribute to resolution of this issue. 3.2.3 Aerosol mixing-state with respect to water uptake A complicated composition mixing state is expected in the ambient atmosphere due to the various primary particle sources and numerous processes altering particle composition. Each kind of internally mixed particle may contain a range of fractional contributions of each constituent, hereinafter referred to as quasi-internal mixtures. External mixing of differently hygroscopic compounds results in distinct growth modes, whereas quasi-internal mixing results in spread of observed growth factors around the growth factor corresponding to the mean composition. This description is consistent with the discussion of composition presented in Sect. 3.1.4.2.1. Detection of different GFD modes or a continuous GFD distribution with a HTDMA implies that the aerosol is to some extent externally mixed. However, detection of a single growth factor does not exclude the presence of externally mixed particles, as different mixtures may still have similar growth factors. At least two distinct growth factor modes are frequently observed in various environments, showing that external mixing is widespread. Some studies even report three to six simultaneous growth modes (Berg et al., 1998a; Cocker et al., 2001; Carrico et al., 2005). In both Aitken and accumulation mode size ranges, an externally mixed mode of marginally hygroscopic particles (GFD¡1.15 at 90% RH) is frequently found in most environments. Numerous studies combining different methods indicate that these marginally hygroscopic particles are dominated by elemental carbon (Covert and Heintzenberg, 1984; Heintzenberg and Covert, 1987; Zhang et al., 1993; Maßling et al., 2003) and/or mineral dust (Zhang et al., 1993; Dick et al., 1998, 2000; Weingartner et al., 2001), as can be expected from the hygroscopic behaviour of pure substances (Table 2). Mineral dust appears to be found in particles with small growth factors only, whereas elemental carbon is also found in particles with larger growth factors. Small amounts of inorganic salts or organic compounds may also be present in these marginally hygroscopic particles being responsible for small water uptake. Though marginally hygroscopic particles are nearly ubiquitous, the dominating growth mode in most environments is readily hygroscopic with GFD values ranging from about 1.2 to 1.8 at 90% RH. Inorganic salts and organic compounds are understood to be the main ingredients of these, quasi-internally mixed, readily hygroscopic www.atmos-chem-phys.net/6/2593/2006/ Atmos. Chem. Phys., 6, 2593–2649, 2006