Cloud albedo increase from carbonaceous aerosol

. Airborne measurements from two consecutive days, analysed with the aid of an aerosol-adiabatic cloud parcel model, are used to study the effect of carbonaceous aerosol particles on the reﬂectivity of sunlight by water clouds. The measurements, including aerosol chemistry, aerosol microphysics, cloud microphysics, cloud gust velocities and cloud light extinction, were made below, in and above stratocumulus over the northwest Atlantic Ocean. On the ﬁrst day, the history of the below-cloud ﬁne particle aerosol was marine and the ﬁne particle sulphate and organic carbon mass concentrations measured at cloud base were 2.4 µg m − 3 and 0.9 µg m − 3 respectively. On


Introduction
Climate prediction is challenged by uncertainties in the cooling from the effect of anthropogenic aerosol particles on the reflectivity of sunlight by water clouds. The effect is most notable for lower altitude clouds of modest thickness such as stratocumulus. This cloud albedo effect (also known as the Twomey effect and as the first indirect effect of aerosols) is rooted in the control of cloud droplet number concentrations (CDNC) by aerosol particles acting as cloud condensation nuclei (CCN) (Twomey, 1977). Twenty-nine estimates of the increase in the global radiative forcing due to the cloud albedo effect range from −0.2 W m −2 to −1.85 W m −2 (Forster et al., 2007), underscoring the significance of this effect and the uncertainty in its prediction. Scaling of general circulation models with satellite data suggests the cloud albedo effect can be constrained as −0.7±0.5 W m −2 (Quaas et al., 2009) and an estimate based on an energy balance approach is consistent with that result (Murphy et al., 2009).
Anthropogenic sources of carbonaceous and sulphate compounds increase the number and mass concentrations of atmospheric aerosol particles and CCN. As a major constituent of the global fine particle aerosol, sulphate components are believed to contribute significantly to the cloud albedo effect due in large part to their relatively high hygroscopicity. Carbonaceous components are also ubiquitous constituents of the global fine particle aerosol Bahadur et al., 2009). Since Novakov and Penner (1993) demonstrated a connection between carbonaceous particles and CCN, many observations have indicated varying degrees of effectiveness of the carbonaceous components at water uptake. Typically carbonaceous components are much less hygroscopic than sulphate components (e.g. McFiggans et al., 2006;Petters and Kreidenweis, 2007) and their effects on CDNC and cloud albedo are highly uncertain. In the absence of significant water activity of the carbonaceous aerosol, a relatively small fraction of sulphate in a particle can influence the ability of the mixed sulphatecarbonaceous particle to act as a CCN (e.g. Lohmann et al., 2004;Petters et al., 2006;King et al., 2007;Prenni et al., 2007). How carbonaceous particles contribute to the aerosol number distribution may represent their most significant CCN influence, since the number distribution is important for the number of CCN (e.g. Russell et al., 1999;Roberts et al., 2002;Dusek et al., 2006;Pierce et al., 2007) and potentially the CDNC (e.g. Conant et al., 2004;Meskhidze et al., 2005;McFiggans et al., 2006). Fountoukis et al. (2007) explicitly considered the role of the carbonaceous aerosol in the nucleation of cloud droplets and found limited sensitivity of their CDNC-aerosol closure study to the solubility of organic mass (OM). On the other hand, ambient CCN observations indicate that the hygroscopicity of the carbonaceous components takes on greater relative importance as the fraction of sulphate is reduced (e.g. Roberts et al., 2002;Ming and Russell, 2004;Chang et al., 2007;Prenni et al., 2007;Wang et al., 2008;Gunthe et al., 2009), and Wex et al. (2009) found an increase in the hygroscopicity of secondary carbonaceous components with increasing aerosol water content. Thus, there will be situations for which the typically lower hygroscopicity of carbonaceous aerosols has a significant influence on the droplet growth rates.
The possibility of carbonaceous material, in the form of organic films, impeding water condensation and effectively lowering the mass accommodation coefficient for water vapour condensing on the growing droplets (α c ) has been considered for several years (e.g. Penner et al., 2001). As droplets grow, albeit more slowly, such films are likely to break, thus removing the impedance to water uptake (e.g. Feingold and Chuang, 2002). Johnson et al. (2005) found that volatilization of a low solubility organic that may have coated particles resulted in an increase in their water uptake rate suggesting suppression by an organic material. Kinetic inhibition has also been observed with some anthropogenic aerosols (Ruehl et al., 2008(Ruehl et al., , 2009Shantz et al., 2010). Shantz et al. (2010) found that values of α c for concentrated solution droplets needed to be in the range 0.04-0.07 in order to explain observed rates of water uptake by anthropogenic particles. As droplets dilute and the situation approaches water on water condensation, a value of unity is appropriate for α c (Wagner et al., 1982;Mozurkewich, 1986;Davidovits et al., 2004;Laaksonen et al., 2005). Particularly with field observations, the effects of mass accommodation, surface tension and other physical-chemical properties of the carbonaceous aerosol are not always easily distinguished.
Global simulations of the indirect effect have begun to move away from empirical prediction of CDNC based on a parameterization of sulphate as a surrogate towards the explicit representation of the major global aerosol components. These mechanistic approaches are required for long range prediction; although they have not yet reduced the uncertainty range of the indirect aerosol effect because the aerosol components and their size distributions introduce new uncertainties (Menon et al., 2003;Penner et al., 2006;Lohmann, 2007;Pringle et al., 2009). Some studies that have used mechanistic treatments of the carbonaceous and sulphate aerosols have attributed more radiative cooling via the cloud albedo effect to the carbonaceous aerosol than to the sulphate aerosol. Lohmann et al. (2000) found that the change in the radiative cooling by the cloud albedo effect between present-day and pre-industrial time ranged from near zero to −0.4 W m −2 due to anthropogenic sulphate and −0.9 W m −2 to −1.3 W m −2 due to anthropogenic carbonaceous aerosol. The reduced effect of the sulphate aerosol relative to the carbonaceous aerosol was a combination of a larger increase in the carbonaceous burden and a larger mode radius of the sulphate. Chuang et al. (2002) estimated a global cloud albedo effect of −1.85 W m −2 , of which −0.30 W m −2 was associated with anthropogenic sulphate, −1.16 W m −2 was associated with carbonaceous aerosols from biomass burning, and −0.52 W m −2 was associated with carbonaceous aerosols from fossil fuel combustion; the total forcing does not equal the sum of the individuals because of the nonlinear relationship between CDNC and aerosol number concentrations. Together, these two studies suggest that the impact on cloud albedo by the carbonaceous aerosol is 3-6 times that due to the sulphate. On the other hand, Ghan et al. (2001) found their mechanistic model indicated the cloud albedo effect was dominated by sulphate.
Here, we present observational evidence to show that the carbonaceous aerosol can indeed enhance the cloud albedo effect as suggested by Lohmann et al. (2000) and Chuang et al. (2002). We contrast airborne observations of aerosol and cloud properties from two flights on back-to-back days over the Atlantic Ocean. These cases are unique in two ways. First, the trajectories suggest a marine character to the aerosol on the first day and a continental/anthropogenic composition to the aerosol on the second day. Second, the number size distribution of the cloud base aerosol sampled on the second day is similar in shape over the range of 0.06-0.8 µm diameter and approximately twice the concentration of that sampled on the first day. The increase in the size distribution is associated with an increase in the submicron carbonaceous aerosol mass concentration relative to the submicron sulphate aerosol mass concentration and the increase is reflected in the observed CDNC and cloud light scattering.  (Draxler, R. R. and Rolph, G. D., 2003;Rolph, 2003) for end points at the approximate time and location of vertical profiles through the clouds: (a) flight 1; (b) flight 2.

Overview
The measurement platform was the National Research Council of Canada Convair 580 aircraft. The two flights considered here were conducted in and around stratocumulus on 13 October 2003 and 14 October 2003 within a few hundred kilometers east of Nova Scotia; hereafter referred to as flights 1 and 2 respectively. The flights were part of the Canadian SOLAS program. Profile data from flights 1 and 2 were collected near 42 • 48 ′ N, 62 • W and 44 • 50 ′ N, 57 • 20 ′ W respectively. The meteorological back trajectory analyses from the NOAA HYSPLIT model (Draxler and Rolph, 2003;Rolph, 2003) indicate that the aerosol sampled during flight 1 had resided over the Atlantic Ocean to the south-southwest of the sampling area for three days prior to observation (Fig. 1a). The trajectory analysis for flight 2 indicates a significant continental influence to the aerosol (Fig. 1b).

Instrumentation
Relevant measurements and the corresponding instrumentation are outlined in Table 1.
Comprehensive descriptions of the Aerodyne Quadruple Aerosol Mass Spectrometer (Q-AMS) are found elsewhere (Jayne et al., 2000;Jimenez et al., 2003). The Q-AMS was used to measure the average mass concentration and size distributions of non-refractory particulate species every 5-min; 5-min averages were chosen to optimize the signal-to-noise versus the spatial resolution. The temperature of the vaporizer in the AMS, used to volatilize the particles into molecular fragments, was set to about 550 • C enabling the measurement of components such as ammonium sulphate and many organics, but not sodium chloride. The transmission of particles into the AMS is approximately 100% in the range 0.1-0.7 µm vacuum aerodynamic diameter (D va ), decreasing sharply outside that range ; the upper transmission limit for this particular AMS has been demonstrated (Rupakheti et al., 2005). The collection efficiency of particles by the oven is assumed to be 100% for this dataset. This assumption is based on the comparison of the sulphate mass concentrations from the Q-AMS with those from the Particle-In-Liquid-Sampler (PILS; described below), and it is consistent with previous results with this Q-AMS (Rupakheti et al., 2005;Buset et al., 2006;Phinney et al., 2006;Langley et al., 2010); more acidic sulphate, as here, is efficiently sampled. Analysis of AMS data was performed using the Deluxe 1.29 IGOR data analysis package (Allan et al., 2003) with a batch file (used for quantitative calibration) and fragmentation file (used for identification of chemical species present on the aerosol) customized to this data set. The fragmentation file included methanesulphonic acid (Langley et al., 2010), but MSA was not found to be significant in these observations. Aerosol particles were collected in water using a PILS and analyzed for their major water soluble inorganic chemical components onboard with two ion chromatographs (IC). The PILS collection system is described briefly here, and Table 1. Instrumentation list.
Cloud liquid water content PMS King Probe -outboard under wing -1 s ave.
Cloud light Extinction Gerber Cloud Integrating Nephelometer (CIN) -outboard under wing -1 s ave. it is similar to that described by Orsini et al. (2003); this particular PILS has been previously documented (Buset et al., 2006). Briefly, the system used two Dionex Ion Chromatographs (ICS 2000) with eluent generation, temperature control and degassing for anions and cations. The anions and cations were eluted in 25 mM Potassium Hydroxide and 25 mM Methanesulfonic Acid respectively. Trace gases were denuded from the aerosol prior to entering the PILS, where the particles are grown to supermicron size droplets under supersaturated conditions created by mixing the sample air with steam. Particles larger than 0.03 µm diameter that enter the PILS are activated with a >97% collection efficiency at a sample flow rate of 15 l m −1 as used here. The droplets are impacted onto a surface that is washed off with a steady stream of de-ionized water. The water is delivered to trace concentrator columns of the IC for analysis. Samples were averaged for 10 min. Four instruments were used to measure the size distributions of the aerosol particles. A TSI Scanning Mobility Particle System (SMPS) with a TSI 3010 Condensation Particle Counter (CPC) was used to measure particles from 10 nm to 500 nm geometric diameter (D g ). A TSI Aerodynamic Particle Sizer (APS) was used to measure particles from 0.6 µm to 20 µm aerodynamic diameter (D a ). Both the SMPS and APS were located inboard the aircraft. Two Particle Measuring Systems (PMS) probes, a Passive Cavity Aerosol Spectrometer Probe (PCASP100x) and a Forward Scattering Spectrometer Probe (FSSP300), were used to measure particles from 0.14 µm to 1 µm D g and from 0.3 µm to 20 µm D g , respectively. The PCASP100x and the FSSP300 were suspended from a pylon under a wing of the aircraft. The FSSP is a non-intrusive probe and the particle sizing includes any water that the particles retain. The other three instruments (SMPS, APS and PCASP) are assumed to have measured the nearly dry sizes of the particles due to internal heating (PCASP), the slight warming of air as it entered the aircraft from outside (SMPS and APS) and dried sheath air used (all three). A comparison of this PCASP100x and FSSP300 is discussed by Strapp et al. (1992). We use the 1 Hz observations from the PCASP to illustrate the changes of particle number concentrations in the profiles as well as the variability during level flight below cloud base. The number concentrations from the PCASP and those from the SMPS for >140 nm agree to within 20%. We do not include the PCASP measurements in the size distributions because the optical sizing is more uncertain. We also note that differences between optical sizing and mobility sizing have been shown to affect the calculated CDNC .
A shrouded nearly isokinetic diffuser set 15 cm off the starboard side of the fuselage ahead of the engine was used to bring the aerosol into the cabin. From there, the aerosol was carried by 1 cm OD stainless steel tubing to the SMPS, APS, AMS and PILS. The distance from the intake point to the SMPS and APS was about 1 m and about 5 m to the AMS and PILS. Agreement between the APS and FSSP300 size distributions was reasonable in the overlap region below 1 µm. Above 1 µm, the particle concentrations measured with the APS were significantly lower than the FSSP300 ( Fig. 2a; the APS aerodynamic diameters were converted to geometric diameter assuming spherical particles and a density of 1.5 g cm −3 ). Drying of the particles may contribute to some of the discrepancy above 1 µm, but this result indicates that losses in the inlet were only significant for particles >1 µm. The measurements from the FSSP300 are used to define the size distribution for particles >0.9 µm. We do not adjust the FSSP300 measurements for water on the particles; Scattering Spectrometer Probe for coarse aerosol particles (FSSP300; 300-20 000 nm) as well as data from the inboard Aerodynamic Particle Sizer distributions (APS; sizes converted to a geometric diameter assuming a density of 1.5 g cm −3 ). Divergence of the FSSP300 and APS above 1 µm is believed to be mostly due to losses at the aircraft intake. (b) Modeled fits to the measured size distributions; details given in Table 5. For (b), we use the SMPS, the APS up to 900 nm and the FSSP300 above 900 nm. the relative humidities measured during level sampling below cloud were 83% and 85% for flights 1 and 2 respectively. We note that the accuracy of the coarse particle measurements is not a significant concern for the present study, but it can be important when considering giant CCN that may influence precipitation formation.
Cloud liquid water contents (LWC) were measured using a PMS King LWC probe and a Nevzorov LWC probe deployed from a pylon suspended from a wing; comparison of the LWC measurements is shown in Fig. 3. The LWC is measured to within 15% and the baseline drift is estimated to be <0.02 g m −3 (Cober et al., 2001). Correction factors applied for the effects of flow around the aircraft were between 1.03 and 1.05 (Cober et al., 2001;Drummond and MacPherson, 1985).
Two PMS FSSP100 probes, also suspended from a pylon under a wing, were used to measure the cloud droplet number size distribution and covered the size range of 2-47 µm. Their sample volumes are based on the measured sample area of the probe and the measured airspeed. Sizing calibrations, corrected to the index of refraction of water, were done with glass beads. The data are 1 s average values, representing a a.  Thirteen minutes were spent directly under cloud base during flight 1 and 6 minutes during flight 2. These periods are also those used to define the chemistry of the cloud base aerosol as given in Table 3. sampling path length of 90-100 m. The cloud droplet number concentrations (CDNC) were corrected for coincidence error and probe dead time (Baumgardner et al., 1985). Corrections of the FSSP100 data for flow around the aircraft were not taken into account because the effects of flow on the measurements (ca. 2%) were significantly less than the probe measurement accuracies for the CDNC (±15%) (Cober et al., 2001). Measurements of the cloud scattering coefficient were made using a Gerber Scientific Inc. Cloud Integrating Nephelometer (CIN; Gerber et al., 2000;Garrett et al., 2001). The CIN probe was suspended from a pylon under a wing. It illuminates a population of cloud particles with laser light at 625 nm wavelength and measures the near-forward and nearbackward scattered light intensities from cloud particles 2 µm and larger. The light scattering coefficients from the CIN probe correspond with the light scattering coefficients calculated from the cloud droplet size distributions measured with the FSSP100 probes (Fig. 3c). The CIN probe provides a measurement that is independent from the FSSP100 probes, thereby avoiding ambiguities that can occur when comparing the CDNC with cloud light extinction.

Observations
Aerosol number concentrations for particles >140 nm (APNC 140 ) and CDNC during profiles through the clouds are shown in Fig. 4a and b; 140 nm roughly corresponds to the lower size of particles that will activate to cloud droplets in these types of clouds (e.g. Leaitch et al., 1996;Russell et al., 1999). The mean APNC 140 at cloud base during flight 1 was 170 cm −3 ± 50 cm −3 and the mean CDNC through the profile was 190 cm −3 ± 50 cm −3 for cloud liquid water contents (LWC) >0.15 g m −3 ; smaller LWC are excluded to remove reductions in the CDNC by evaporation at cloud edges. For flight 2, the mean APNC 140 for cloud base was 390 cm −3 ±100 cm −3 and the mean CDNC through the profile was 560 cm −3 ±110 cm −3 . The cloud base aerosol and CDNC sampled during flight 2 are higher than those of flight 1 at a confidence level of >99% as in Table 2. The time series of the observations are shown in Fig. 5.
The LWC profiles were near adiabatic and the steady increase of the LWC from bottom to top indicates lifting from cloud base (Fig. 3); detection of an aerosol-cloud albedo effect is easier for nearly adiabatic clouds (Kim et al., 2008). The cloud light scattering values were higher for the profile of flight 2 (Fig. 3c) Observations are for the periods when cloud was profiled and sampling was conducted above, in and below cloud at level intervals. The CDNC are from the two FSSP100 probes (serial numbers 096 and 124) corrected for coincidence and probe dead times (Baumgardner et al., 1985), and the APNC >140 is from the PCASP100X. Table 3. Aerosol Chemistry at Cloud Base. The mass concentrations of the chemical compounds measured at cloud base with the Q-AMS and the PILS (the PILS values are in parentheses) for the times given in Table 2. The PILS average from flight 2 is shown as a range of two samples that cover segments of the five-minute Q-AMS average. The Q-AMS non-refractory mass concentrations are slightly different than in the profiles of Fig. 7   droplet size distributions from one of the FSSP100 probes (096) averaged through each profile are shown in Fig. 6. The mode diameter is about 12 µm for the distribution of flight 1 and about 8 µm for the flight 2 distribution. Although momentarily flying out of cloud contributes significantly to the lower LWP of flight 2, we note that recent studies have associated increases in aerosol with slight reductions in LWP (Avey et al., 2007;Lebsock et al., 2008). Profile data of the non-refractory chemical composition of the cloud base fine particle aerosol is shown in Fig. 7. The fine aerosol was dominated by sulphate components on flight 1 and by a mix of organic components (OM: carbonaceous material exclusive of elemental carbon) and sulphate components on flight 2; respective fine particle sulphate and OM mass concentrations measured for level flight at cloud base were 2.4 µg m −3 and 0.9 µg m −3 during flight 1 and 2.3 µg m −3 and 2.6 µg m −3 during flight 2 (Table 3). Collection efficiency aside, the maximum uncertainty for the mass concentrations measured with this Q-AMS is estimated at ±20% of the mass concentrations (Rupakheti et al., 2005).
For the range 0.06-0.8 µm, the aerosol particle size distributions at cloud base ( Fig. 2) differ principally in number concentration; the flight 2 cloud base number concentration is about twice that of flight 1 above 0.06 µm. The variations of the mass fractions of OM relative to sulphate plus OM with particle size from 0.1 µm to 0.5 µm are shown in Fig. 8a. The particle volume concentrations estimated from the sums of sulphate and OM compared with the total volume distributions estimated from the physical size distributions are shown in Fig. 8b; volumes were calculated from the AMS assuming densities of 1.78 g cm −3 for sulphate and 1.2 g cm −3 for OM. No other chemical species of significant concentration were measured in this size range with the Q-AMS. Mass concentrations of sodium and chloride ions measured with the PILS were associated with coarse particle number concetrations (Fig. 8c). We assume that the discrepancies in the volume concentrations above 0.5 µm (Fig. 8b) are explained by sea salt components. Table 4  The time series of vertical gusts or updraft speeds are shown in Fig. 9. We sample the updrafts across the maximum available level-flight path in cloud. For a sufficient path length in this type of cloud, the mean of the gusts is zero by definition. The variations in updraft speeds represent the potential influence on the CDNC. We use one standard deviation of the vertical gust measurements during level flight, as discussed by Peng et al. (2005), to define the variation. One standard deviation is 14 cm s −1 for flight 1 and 50 cm s −1 for flight 2. The maximum gusts were approximately 20 cm s −1 for flight 1 and 100 cm s −1 for flight 2. The relative increase in particles >2000 nm at lower altitudes, the association of sodium with increased concentrations of these larger particles and the higher winds speeds of flight 2 suggest the presence of sea salt particles.

Modelling
The aerosol-cloud adiabatic parcel model includes representation of weakly soluble compounds (Shantz et al., 2003) as well as the effects of different surface tensions (Lohmann et al., 2004). The cloud base aerosol size distributions are described by multiple modes each represented as a log-normal function (Fig. 2b). The chemical representations of each mode are based on the AMS and PILS data for particles >0.1 µm. Smaller particles are assumed to be either 100% H 2 SO 4 or 70% organic and 30% sulphate by mass, as indicated. Table 5 gives the physical and chemical representations for each mode used in the simulations; other initial conditions used in the simulations are given at the bottom of Table 5. Except as discussed, all chemical components within each mode are assumed to be internally mixed. The -Accommodation coefficient for water, 1.0; sensitivity to 0.06.
simulated CDNC are defined as droplets >2 µm diameter that are still growing after the maximum supersaturation is reached, in accordance with the lower size of the FSSP100 and the measured droplet distributions (Fig. 6). The model uses a cloud updraft speed to define the rate of cooling of the air parcel that eventually leads to a water supersaturation and nucleation of cloud droplets; higher updraft speeds increase the cloud base supersaturation and the resulting droplet growth rates. The method of assessing the appropriate updraft speed to use with an adiabatic parcel model is not well established. In the case of stratiform cloud, as in the present study, it is difficult if not impossible to identify an updraft core. We use the results of the Peng et al. (2005) analysis that indicated a value of one standard deviation of the measured gust velocities provided the best closure of the aerosol and cloud measurements. The physical basis for this is that the smaller positive updrafts sampled when flying level through a stratiform cloud are the result of changes in the stronger vertical motions within cloud, and they are not the updrafts that penetrate from below cloud base up into the cloud. In other words, there is some distribution of higher updrafts that specify CDNC in the cloud. Based on this reasoning, we expect the updraft speeds higher than one standard deviation are also a factor in determining the CDNC. As discussed above, we use updraft speeds of 14 cm s −1 for flight 1 and 50 cm s −1 for flight 2 as principal model inputs. We also conduct simulations with the maximum gusts: 20 cm s −1 for flight 1 and 100 cm s −1 for flight 2. As discussed below, we also apply the updraft speeds of flight 2 to the aerosol of flight 1.
The sensitivity of the simulated CDNC to the assumption of the composition of the smallest particles that can contribute to the CDNC in this case (mode 2 in Table 5) was considered in the case of flight 2 by changing the composition from H 2 SO 4 to 70% OM and 30% H 2 SO 4 assuming a solubility of 5 g l −1 for the OM. The sensitivity to the assumption of an internal mix of the organics and the sulphate was tested by assuming that the OM and the sulphate in modes 2 and 3 of flight 2 were externally mixed. These results are shown under "External mixture" in Table 6. The surface tension used in these calculations is based on that of adipic acid (Ervens et al., 2004). The impact of this surface tension is considered by simulating the flight 2 internally mixed cases for a surface tension of water plus ammonium sulphate for the OM solubilities of 5 and 200 g l −1 and for updrafts of 50 and 100 cm s −1 . These cases were selected because they are the conditions closest to the observations and should offer the greatest sensitivity to the surface tension due to the higher organic concentrations in solution.
We use a value of unity for α c in the simulations, assuming that the droplets by the time they reach cloud base are sufficiently dilute that α c = 1 is appropriate. We also test the effect of defining α c as 0.06 as used by Conant et al. (2004) and by Fountoukis et al. (2007).
The cloud albedo (A c ) is calculated using where τ is the optical depth calculated from the integration of the extinction measurements through the cloud profile (e.g. Menon et al., 2003). The planetary albedo (A p ) using where in this case A s is the underlying reflectance from the ocean surface and assumed to be 0.1. The local radiative forcing is simply estimated assuming an increase of −2 W m −2 for a planetary albedo increase of 0.5% (Ramanathan, 1988).

Results and discussion
The CDNC simulated for flight 1 assuming internally mixed sulphate and OM (225 cm −3 ; Table 6, 14 cm s −1 ) are within the range of the observed values (140-240 cm −3 ) and within 18% of the observed mean CDNC. There is little or no change in the CDNC with change in the OM solubility due to the relatively large sulphate fraction that controls water uptake. For flight 2, the CDNC was simulated for the three possible unique states of the cloud base aerosol. First, for the assumption that the carbonaceous components of flight 2 were particles externally mixed relative to the sulphate components and either hydrophobic or weakly hygroscopic (Fig. 10a) the simulated CDNC (370 cm −3 ; Table 6) is outside of the range of the observed CDNC (450-770 cm −3 ) and 34% lower than the mean. Second, for the assumption that the carbonaceous particles were externally mixed and highly hygroscopic (Fig. 10b) the resulting CDNC (560 cm −3 ; Table 6) is equivalent to the mean of the observed CDNC. Third, for the assumption that carbonaceous particles served primarily as substrates for sulphate condensation (internal mixture, Fig. 10c) and defined the CCN number concentration the simulated CDNC (491-579 cm −3 ; Table 6) are within the range of the observed CDNC and the mean (535 cm −3 ) is within 7% of the observed mean. Reducing the surface tension in these models produced no significant change in the CDNC, and reducing the mass accommodation coefficient to 0.06 (e.g. Fountoukis et al., 2007) increased the CDNC by 12%. The only aerosol model of flight 2 that does not represent the increase in the CDNC of flight 2 relative to flight 1 is that which assumes no effect of the carbonaceous particles on the CDNC (Fig. 10a). Thus, we find that the carbonaceous components of the below-cloud aerosol contributed to the increase in the CDNC.
The updraft speed of flight 2 is applied to the flight 1 aerosol in order to assess the degree to which the higher updrafts of flight 2 were responsible for the higher CDNC. The resulting increase in the CDNC (330 cm −3 from 225 cm −3 ; Table 6) is about 30% of the difference between flights 1 and 2 (560 cm −3 from 225 cm −3 ), leaving 70% of the increase attributable to the change in the aerosol. For 70% of the difference in the optical depths (4.2-3.8), we estimate that the increase in aerosol due to the carbonaceous components produced a local increase in cloud albedo of 5.5%, which equates to a local radiative forcing of approximately −12 W m −2 for an ocean surface reflectance of 10%. Any additional absorption effects from the increase in carbonaceous components of flight 2 are not considered in this estimate, while the absolute value of the forcing is underestimated due to the reduced vertically integrated cloud liquid water of flight 2.
The above attribution is predicated on the higher updraft speed during flight 2 as the application of the weaker updraft of flight 1 to the aerosol of flight 2 simulates a CDNC that is slightly lower than that of flight 1. A lower updraft slows the rate of condensation, and only CCN with larger critical diameters will activate or reach cloud droplet size. In The carbonaceous (green) and sulphate (red) components are externally mixed with no water uptake by the carbonaceous particles; (b) externally mixed and the carbonaceous particles strongly influence the water uptake; (c) internally mixed carbonaceous and sulphate components with no effect of the carbonaceous components on water uptake. that situation, the chemistry has a substantial bearing on the lower activation diameter; CCN observations that point to the importance of the size distribution do not consider supersaturations <0.1% that are relevant to flight 1. At higher updraft speeds, the cloud base supersaturation is higher (up to 0.2% as modelled for flight 2; Fig. 11), the critical diameter of activation is smaller and the number distribution becomes a stronger influence for the CDNC; however the chemistry of the larger CCN still is important due to its control of the cloud base supersaturation via its effect on the water uptake rate. This point is illustrated here by the decrease in the CDNC for the assumption that the aerosol of flight 2 is all sulphate and the updraft speed of 100 cm s −1 or greater. It is  Fig. 11. Examples of the simulated profiles of relative humidity just below and above cloud base. The air is supersaturated with respect to water where the relative humidity exceeds 100%. Cases of weak solubility of the OM (5 g l −1 ) and pure H 2 SO 4 are shown for both flights 1 and 2. The water supersaturation decreases when the aerosol is all sulphuric acid due to the more rapid rate of water uptake by this chemical component. also consistent with the suggestion that the assumption of organic compounds as hygroscopic affects the sensitivity of the CDNC to the vertical velocity Lance et al., 2004).
Since the flight 2 observations were made at least one day or more downwind of significant carbonaceous aerosol sources (Fig. 1b) the sulphate and carbonaceous components were internally mixed to some degree (e.g., Covert and Heintzenberg, 1984;Heald et al., 2005). Also, previous observations of the residuals of cloud droplets from marine clouds have found that both components co-exist in cloud water (e.g., Straub et al., 2007;Targino et al., 2007;Hawkins et al., 2008). This is the situation represented in Fig. 10c, in which case the ability of the carbonaceous components to affect the water uptake rate of the flight 2 aerosol has a relatively small effect on the simulated CDNC (Table 6) due to the influence of the sulphate. However, if we assume that the particles are composed completely of sulphate a slight reduction in the CDNC results because the large increase in droplet growth rates reduces the cloud base supersaturation ( Figure 10); despite the slightly lower value, the CDNC resulting from the all sulphate scenario falls within the range of the observed CDNC.
The effects of carbonaceous components on the CDNC are inherent in the sulphate-CDNC empirical parameterization (Penner et al., 2001), but they can not be separated from those of sulphate. Menon et al. (2002) developed an empirical relationship for CDNC in an attempt to explicitly identify changes in the CDNC due to changes in sulphate, organic, and sea salt mass concentrations. Application of that for-mulation to the present observations produces only a 15% increase in the CDNC from flight 1 to flight 2 compared with the 70% increase based on the present observations and analysis. Pringle et al. (2009) finds no reason to suggest mechanistic models are currently better than empirical models, other than the potential for the mechanistic model to better represent regional effects. However, as we begin to consider the effects of changes in emissions (e.g. Kloster et al., 2008), our result emphasises the point that accurate distinction between the carbonaceous and sulphur components is essential for assessing changes in the cloud albedo effect, and continued development of mechanistic approaches is the only way to achieve the needed confidence in our ability to predict effects of emissions changes.

Summary and conclusions
We compare measurements made above, in and below stratocumulus during flights on two sequential days over the Atlantic Ocean to consider the effect of carbonaceous aerosol particles on the reflectivity of sunlight by water clouds. On the first day, the below-cloud fine particle aerosol was mostly of marine character and the cloud base fine particle sulphate and organic carbon mass concentrations measured with a Q-AMS were 2.4 µg m −3 and 0.9 µg m −3 respectively. On the second day, the below-cloud aerosol was also continentally influenced and the sulphate and organic carbon mass concentrations from the Q-AMS were 2.3 µg m −3 and 2.6 µg m −3 respectively. Correspondingly, the number concentrations of aerosol particles below cloud were approximately a factor of two higher on the second day, while the shapes of the below-cloud size distributions over 0.06-0.8 µm diameter were similar on both days. The CDNC on the second day were approximately three times higher than the CDNC measured on the first day. In addition, the measured cloud light scattering was higher on the second day despite a slightly lower liquid water content. The vertically integrated cloud light extinctions indicate an increase in the cloud albedo from the first day to the second day of approximately 8%. Using an aerosol-adiabatic cloud parcel model to separate the influence of the differences in the cloud gust velocities, we estimate that fractionally 0.7 of the 8% increase in cloud albedo on the second day was due to the higher concentrations of below-cloud aerosol that principally resulted from the increase in the carbonaceous components of that aerosol. Assuming no additional absorption by this aerosol, the estimated local albedo increase due the carbonaceous components in the aerosol translates to a daytime radiative cooling of ∼ 12 W m −2 .
Although we only contrast two cases, the result suggests that the contribution to radiative forcing via the cloud albedo effect from carbonaceous components can be not only substantial but significantly higher than the contribution from sulphate, in agreement with some global models using mechanistic approaches to represent the aerosol and CDNC.