Correlation between cloud condensation nuclei concentration and aerosol optical thickness in remote and polluted regions

Introduction Conclusions References Tables Figures


Introduction
The concentration of cloud-active particles, especially in the lower troposphere, has a profound influence on the microphysical processes in clouds, and consequently on many aspects of weather and climate.These interactions have been summarized in a number of recent reviews, addressing in particular the effects of aerosols on climate (Penner et al., 2001;Lohmann and Feichter, 2005;IPCC, 2007) and on cloud processes and precipitation (McFiggans et al., 2006;Rosenfeld, 2006a;IAPSAG, 2007;Andreae and Rosenfeld, 2008).In addition to their cloud microphysical effects, aerosols also modulate cloud formation and convective behavior through their radiative effects, for which aerosol optical thickness (AOT) is a commonly used metric.AOT has the advantage of being readily observed by remote sensing, and AOT measurements are now done routinely from space by several sensors (Kaufman et al., 2002;Yu et al., 2003;Kahn et al., 2007;Kokhanovsky et al., 2007) as well as by ground-based sunphotometer networks (Holben et al., 2001;Kim et al., 2008).
In order to incorporate the effects of cloud condensation nuclei (CCN) in meteorological models at all scales, from large eddy simulation (LES) to global climate models (GCM), knowledge of the spatial and temporal distribution of CCN in the atmosphere is essential.This information is, however, difficult to obtain from observations.In-situ measurements of CCN concentrations only provide very localized and sparse information, while the detection of CCN by remote sensing has not yet been accomplished.This is due to the difference in size ranges important for CCN concentrations on one hand, and for light extinction on the other.
The ability of a particle to nucleate a cloud droplet depends on its size and composition.The latter is now frequently represented by the hygroscopicity factor, κ, which typically falls in the range of 0.1-0.9 for ambient aerosols (Petters and Kreidenweis, 2007;Andreae and Rosenfeld, 2008).At these values of κ, particles must have diameters larger than about 40-70 nm in order to activate at the highest supersaturations commonly found in clouds (up to about 0.6%).This therefore represents the lower boundary of the size range containing CCN-active particles.The upper boundary is effectively defined by the sharp decrease in the number concentration of particles as a function of size, so that usually only a minor fraction of CCN is in the size range above some 200-300 nm diameter (Seinfeld and Pandis, 1998).The maximum of the CCN size distribution thus typically falls in the range of about 70-200 nm.This range is well below the maximum in the Mie scattering efficiency function for light with a wavelength of about 500 nm, which is most commonly used to represent the aerosol optical thickness (AOT) of the atmospheric column.In contrast, the maximum of the scattering and extinction efficiency functions often falls near or above the maximum of the mass size distribution of the aerosol in the range between 400 and 1000 nm, so that this part of the size distribution usually has the strongest influence on the AOT 500 values.In regions with high loadings of dust and seasalt aerosol, the coarse mode (>1 µm diameter) may also contribute strongly to AOT 500 (Mulcahy et al., 2008;Remer et al., 2008).
The disconnect between the parts of the aerosol size spectrum dominating the CCN abundance and those dominating visible light extinction suggests that correlations between these two variables may not be very strong.Consequently, the use of remote sensing measurements for the estimation of CCN abundances has been considered difficult (Gasso and Hegg, 2003;Ghan et al., 2006;Kapustin et al., 2006;Rosenfeld, 2006b).Further complications arise from the fact that column-integrated properties, such as AOT, are not necessarily representative of surface or near-surface properties such as the concentration of aerosol particles and CCN in the boundary layer.Variability in the scale height of the vertical aerosol distribution and the existence of aerosol layers aloft can introduce substantial variability in the relationship between column and surface properties.Changes in RH can result in pronounced variations in AOT even while dry aerosol concentrations and CCN concentrations remain the same, adding further variability into a potential relationship between AOT and boundary-layer CCN concentrations.
The motivation for the present study came from a discussion about the choice of variables that would be appropriate as drivers in global models of aerosol effects on clouds and climate, and the ranges of values that would need to be considered for these variables.We selected AOT as a proxy for the radiative effect of aerosols on cloud forma-tion and the production of convectively available potential energy (CAPE) at the surface.CCN 0.4 was chosen to represent microphysical effects.As no global compilation of the available data for these properties was available, and because a substantial amount of new data for both variables has recently become available, I decided to make an empirical investigation of their correlation in collocated (or at least nearly collocated) data sets.A conceptual analysis of the combined radiative and microphysical effects of aerosols on clouds, based in part on the results reported here, has recently been published (Rosenfeld et al., 2008).

Methods
Most AOT values were obtained from the AERONET database publicly available on the Internet at http://aeronet.gsfc.nasa.gov/.These data have the advantage of being available from a global network of stations with consistent processing algorithms and quality control (Dubovik et al., 2000).In most cases, I was able to use Level 2.0 data, which are fully calibrated and cloud-screened.In some instances, only Level 1.5 data were available, which have been cloudscreened, but do not yet have the final calibration applied.These cases, and the few cases where other sunphotometer data were used, have been indicated in Tables 1 and 2. I calculated the statistics reported in Tables 1 and 2 based on the daily average data reported in the AERONET database, in order to avoid introducing bias from variability in the number of measurements available on individual days.I chose not to use MODIS AOT data for this analysis because of persisting uncertainties regarding the absolute accuracy of this data (Li et al., 2007).
I chose to use AOT values at 500 nm (AOT 500 ), because they are most commonly available and most frequently used as a metric for aerosol burdens.When only measurements at other wavelengths were available, I chose the nearest available wavelength and made an adjustment using the appropriate Ångstrom exponent.
To represent CCN concentrations, I used the set of particles that activate at a supersaturation of 0.4%, a value commonly used for convective clouds.Many of the data were taken from the literature, and have been obtained using a variety of instruments about which details can be found in the original papers referenced in the Tables.Our own previously unpublished measurements were obtained with the static chamber counter described by Frank et al. (2007) and the DMT continuous-flow counter using the techniques discussed in Rose et al. (2008).When only data at supersaturations other than 0.4% were available, they were adjusted to 0.4% using the conventional power law formulation for the dependence of CCN concentrations on supersaturation ([CCN S ]=[CCN 1.0 ]*S k ).The exponent, k, was either derived from the data, or where this was not possible, I used default values of 0.5 for continental and 0.4 for marine sites.In order to increase the available database, I deduced CCN concentrations from measurements of aerosol size spectra for some cases where direct CCN measurements were not available.For this purpose, I used the lower cutoff diameters specified in the Comment column in Table 2, chosen based either on measurements of these cutoff diameters from the same site or from similar locations, or obtained using appropriate values of κ, and the relationships between κ and the activation diameter given by Petters and Kreidenweis (2007).Obviously, this introduces additional uncertainty into our estimates, which is in each case a function of the accuracy of the estimate of the cutoff diameter and the shape of the size distribution.In unfavorable cases, this uncertainty may be large (up to about 50%).In most cases, however, the maximum of the size distributions at our rural and remote stations was well above the range of possible cutoff diameters (50-90 nm for the range of κ and supersaturations considered here), and therefore the resulting uncertainty is likely not greater than what one must accept when using literature data from a great variety of groups and instruments.Since we are looking only for fairly broad relationships and consider a wide range of datasets and conditions, it is unlikely that this approach would introduce a systematic bias.
In addition to data sets where corresponding pairs of AOT and CCN data were available, some values where only one of the variables was measured have been added to the Tables for information purposes.Furthermore, condensation nuclei (CN) concentrations have been provided in the Tables, when available.These data have been obtained by a variety of instruments, with different lower cut-off diameters, and thus some caution must be exercised when comparing their values.

Results and discussion
For this study, I have separated the available data into four general regimes, continental-remote, continental-polluted, marine-remote and marine-polluted.I have excluded the analysis of dust-dominated regions downwind of the major dust source regions from our analysis, because there is little or no collocated CCN and AOT data available, and because dominance of the coarse mode in these regions precludes a meaningful relationship between CCN and AOT.
An emphasis on looking at remote regions came from our interest in estimating the pre-human aerosol loading of the www.atmos-chem-phys.net/9/543/2009/Atmos.Chem.Phys., 9, 543-556, 2009 a median b AERONET level 1.5 data atmosphere, which is of relevance to understanding the human perturbation (Andreae, 2007a).At the present time, the most remote, and therefore probably closest to pristine, continental regions are found in Amazonia and in parts of northern North America and Siberia.Because of the seasonal heavy pollution of the Amazonian atmosphere with biomass smoke, this area also provides the opportunity to study very clean and highly polluted conditions within the same region.

Amazonia
Even in an area as remote as the Amazon Basin, the presentday aerosol population is influenced by anthropogenic emissions, which are mostly from biomass burning.When the emission sources are large, as during the regional fire season, or located nearby, this influence can be readily identi-fied.On the other hand, emissions arriving by long-range transport, especially from Africa, can have a significant influence on the aerosol population over Amazonia, even when there is little fire activity in the Amazon Basin (Prospero et al., 1981;Talbot et al., 1990;Swap et al., 1992;Formenti et al., 2001).For this reason, the CCN and CN concentrations summarized in Table 1 must always be considered as upper limits for the pristine values.The first CCN measurements in Amazonia were obtained during CLAIRE-98 by Roberts et al. (2001) at Balbina near Manaus.Subsequently, CCN and CN measurements were made as part of the LBA-EUSTACH and SMOCC programs (Andreae et al., 2002;Andreae et al., 2004).In the course of a thorough investigation of the CCN counter used in the study of Roberts et al. (2001), it was found that the supersaturations in that study had been overestimated (Frank et al., 2007).Therefore, the values were a median b AERONET level 1.5 data corrected to a critical activation diameter of 85 nm at a supersaturation of 0.4%, corresponding to a CCN 0.4 /CN ratio for the CLAIRE-98 data of 0.41.This critical diameter appears to be very robust for the clean Amazonian aerosol, and has been found independently in several studies (Rissler et al., 2004;D. Rose, personal communication, 2008).It has therefore also been used, where necessary, for the calculation of CCN 0.4 values from CN values in Table 1.
The results show surprisingly little difference between wet and dry season measurements in clean conditions.The CLAIRE-98, EUSTACH and AMAZE-08 measurements during the full wet season (190±130, 155±65 and 138±94 cm −3 , respectively) are only moderately lower than the SMOCC-2002 values collected during the full dry season over the western Amazon (205±40 cm −3 ).One can therefore conclude that the natural CCN 0.4 concentrations over Amazonia are centered around a value no greater than 180 cm −3 , with a range of about 80-250 cm −3 , and with only a modest seasonal range of about 40 cm −3 .Again, I emphasize that even these low values contain some anthropogenic contamination from long-range transport, as shown by the presence of soot particles and a slightly elevated sulfate content that could be related to trans-Atlantic transport (Andreae et al., 1990 and unpublished aerosol composition and SEM data from the campaigns listed above).
During the burning season, pyrogenic aerosols from tens of thousands of fires overwhelm the natural aerosol population.It is not very meaningful to give average concentrations for this situation, as the actual concentration at any given time and place is dictated by the proximity to the fires and the meteorological conditions, and therefore varies over orders of magnitude from near-pristine values to those inside fresh plumes, which can be in the hundreds of thousands per cm −3 .Table 1 lists the ranges of typical concentrations observed during several campaigns in the smoky season.One finds that at sites remote from the fires, such as Balbina in northern Amazonas State, CCN 0.4 concentrations in the dry season typically reach up to ca. 1000 cm −3 , while in the heavily impacted states of Mato Grosso and Rondonia, typical concentrations are in the range of 1000-4000 cm −3 .
Aerosol optical thickness measurements over the Amazon Basin from the AERONET network have been reviewed by Schafer et al. (2008).In Table 1 I give the averages from the cleanest 3 months (April-Juni in the northern part of the Amazon Basin, February-April in the southern part) to represent clean conditions, and the average of August and September in the southern part of the Amazon forest to represent the smoky period.The average values for the clean period fall near 0.09, while in the smoky period they range around 0.90, and thus show about the same factor-of-ten increase from clean to smoky conditions as the CCN concentrations.During some of the cleanest episodes, values as low as 0.05 were observed.The Ångstrom exponent, å, during the clean periods is relatively low (0.7-1.1), indicating that coarse particles (primary biogenic material, but also some dust from long-range transport) contribute significantly to AOT.The results from the Amazon forest are nearly identical to values observed at tropical and subtropical marine sites (cf.Table 2 and Fig. 1), such as Lanai (Hawaii) and San Nicolas Island (California), where AOT 500 values of 0.08±0.03have been observed, with å in the range of 0.6 to 1.3 (Holben et al., 2001).
During the smoky period, å increases to 1.7±0.1 as a result of the increased importance of the fine mode aerosol.The single scattering albedo of the smoke aerosol over the Amazon forest is 0.92±0.01,with excellent agreement between sunphotometer and in-situ measurements (Schmid et al., 2006;Schafer et al., 2008).

Remote temperate continental regions
Because the large expanses of temperate ecosystems fall into the same latitude belt as the regions with the highest density of anthropogenic emissions, it is very difficult to estimate pristine CCN concentrations and AOT over them.There is a surprisingly small number of data sets from remote regions in the temperate zone, and none where CCN and AOT measurements are truly collocated.Furthermore, the direct CCN measurements must either be taken from older studies, with sometimes uncertain measurement accuracy, or be deduced from size distributions.In spite of these problems, a surprisingly consistent picture emerges.In western and northern North America, remote sites tend to have CCN 0.4 concentrations ranging from about 90 in winter to ca. 280 in summer.Average AOT 500 values typically fall in the 0.06-0.12range.The northern European CCN concentrations tend to be somewhat higher, which is not altogether surprising considering the likely impact of residual air pollution in this region (Putaud et al., 2004;Van Dingenen et al., 2004).Unfortunately, no AOT data are available from this region.
An area of great potential interest for aerosol studies are the remote regions of Siberia, where the first measurements with modern aerosol size spectrometers have recently been reported (Heintzenberg et al., 2008).A preliminary analysis of a one-year data set from the ZOTTO tall tower site near 60 • N, 90 • E shows an average CCN 0.4 concentration (calculated from CN>60 nm) of about 190 cm −3 during clean periods (defined as those times when the aerosol absorption coefficient is <2 Mm −1 , corresponding to an equivalent black carbon concentration of about <0.2 µg m −3 ).The data show that CN and light absorption by soot particles are still correlated even at low concentrations, indicating the presence of a residual pollution component.It is difficult to find corresponding AOT data, as the two nearest AERONET sites are in Krasnoyarsk and Tomsk, two highly industrialized cities with strong local pollution sources.The median and quartile ranges for AOT 500 at these sites are 0.163(0.122,0.219) and 0.138(0.094,0.209), respectively.A better site to represent remote values in boreal Siberia may be Yakutsk, which has a much smaller population and little industry.The median AOT 500 there is 0.081(0.053-0.120),comparable to many other remote continental sites, but still considerably higher than at Fort McMurray in northern Canada: 0.057(0.034,0.089).The only report from the extratropical continental Southern Hemisphere is from a flight campaign in South Africa, where CCN 0.4 values of 137±63 were measured over the Highveld region on a clean day (Ross et al., 2003).On the same day, the AERONET site at Bethlehem, in the center of the Highveld, measured an AOT 500 of 0.045±0.013.In summary, the mean values of CCN 0.4 and AOT 500 over extratropical remote sites are not distinctly different from those measured over Amazonia, even considering that some influence from long-range transport of pollution aerosol is unavoidable in these measurements.For example, more than half of the sulfate aerosol over remote British Columbia is from East Asian sources (van Donkelaar et al., 2008).This implies that pre-anthropogenic CCN 0.4 concentrations over most continental regions were below, maybe even well below, 200 cm −3 .

Remote marine regions
Remote marine regions, especially in the Southern Hemisphere, are usually considered the least polluted and most pristine parts of the atmosphere, but it must be remembered that anthropogenic pollution reaches even the remotest sites.This is, for example, readily seen at the Cape Grim background station, where aerosols from biomass burning in Southern Africa are readily detected during the fire season (Heintzenberg and Bigg, 1990).Nevertheless, remote marine sites show the lowest number concentrations of aerosol particles and CCN worldwide.In the winter season, CCN concentrations in some areas drop down to a few tens per cm 3 , as shown in Table 2 by data from the Southern Ocean (Cape Grim) and the Northeast Pacific (off Washington State), while in other areas, e.g., at the Northeast Atlantic coast (Mace Head) this seasonal cycle is less pronounced.Over biologically productive ocean regions, such as the tropical oceans and the temperate regions in summer, CCN 0.4 concentrations are typically near or above 100 cm −3 .This seasonal behavior is related to the biogenic production of marine aerosols, probably both via the emission of DMS and its oxidation to sulfate, and the release of primary biogenic particles (Charlson et al., 1987;Andreae and Rosenfeld, 2008).The reduced seasonality at some sites may be related to the effect of low levels of anthropogenic pollution, which can make a significant contribution to the very low aerosol concentrations present at remote oceanic sites in winter (Andreae et al., 1999(Andreae et al., , 2003;;Reade et al., 2006).
Overall, one finds that CCN concentrations over the present-day remote oceans are on average about one-half of those over the present-day remote continents (Table 2), but with a very broad overlap (Fig. 1).In view of the fact that the sources of anthropogenic emissions are all located on land (with the exception of ship-stack emissions), it must be assumed that this ocean-land difference in CCN concentrations was substantially lower in pre-human times.
Remote ocean areas also show very low AOT values, in spite of the relatively high fraction of scattering associated with the seasalt aerosol (Quinn and Coffman, 1999), and the high AOTs reported in one study that included very high wind speeds (Mulcahy et al., 2008).The datasets compiled in Table 2 yield an average AOT 500 of 0.055±0.023, in good agreement with the "baseline" marine AOT 550 of 0.06±0.01given by Kaufman et al. (2005).Interestingly, the seasonality of AOT at Cape Grim is opposite to that of seasalt aerosol, which has its highest concentrations in the winter (Andreae et al., 1999).This argues against a dominant role of seasalt in controlling AOT, and suggests that the fine aerosol fraction may dominate both CCN concentration and AOT.The same conclusion was reached by Vallina et al. (2006) based on a statistical analysis of the CCN concentrations over the Southern Ocean and their controlling variables.

Polluted continental and marine regions
CCN concentrations and AOT values in polluted regions show a continuous increase from near-pristine values to extremely high levels in urban or biomass-burning regions (Table 2).In general, polluted marine regions tend to have lower values (range of study averages: CCN 0.4 600-1700 cm −3 , AOT 500 0.15-0.39)than continental polluted areas (CCN 0.4 370-9100 cm −3 , AOT 500 0.10-0.90)because they are usually more distant from sources.This can also be seen at some coastal sites, such as Amnyeon and Gosan (Korea), which experience both direct continental flow and inflow of polluted airmasses that have spent up to several days over the ocean.Thus, the CCN 0.4 concentrations in these airmasses, which in terms of airmass trajectories would be classified as marine, are in the range conventionally thought of as "continental", again putting in question the validity of this classification (Roberts et al., 2001).On the other hand, some of the continental sites in North America and Europe (e.g., the New Hampshire, North Carolina, and Hohenpeissenberg sites) show relatively low CCN and AOT values, most likely as a result of the reductions in pollutant emissions over the last two decades.The highest values in Table 2 come from peri-urban regions in China, i.e., from locations just outside the urban areas of Guangzhou and Beijing.I have not included any urban measurements, because it is difficult to obtain representative measurements in such a highly variable environment, and because this analysis is mainly directed toward the regional to global scale.
8 Relationship between CCN 0.4 and AOT 500 The scatterplot between CCN 0.4 and AOT 500 (Fig. 1) shows a surprisingly tight relationship, which can be fitted with a power law AOT 500 =0.0027•[CCN 0.4 ] 0.640 with a very high degree of correlation (r 2 =0.88).While the deviations of individual studies from this trend are sometimes large (up to a factor of three), and obviously deviations for single measurements at one place and time must be expected to be even greater, this does provide a basis for a parameterization of CCN concentrations in large-scale regional and global climate models.Note that in almost all cases the regression line goes through the error bars of the data.Figure 1 also highlights the broad overlap between remote marine and continental values.
Some further developments to this approach suggest themselves.Given that the CCN concentration is more closely tied to the finer fraction of the aerosol, AOT measurements at lower wavelengths might provide better correlations than the commonly used AOT 500 , which was employed in this study.Alternatively, instead of AOT 500 , one might examine correlations between CCN and the aerosol index AI, defined as the product of AOT and the Ångstrom exponent, thus providing another way of weighting the AOT measurement toward the fine mode.Finally, instead of the use of ground-based AOT measurements, one could examine the use of products based on satellite remote sensing, such as the fine mode AOT product from MODIS (Remer et al., 2005), especially once the remaining calibration issues in the MODIS products have been resolved (Remer et al., 2008).These investigations go beyond the scope of the present study, however, which has been designed as a first examination of large-scale relationships between potential proxies for the radiative and cloud microphysical forcings of anthropogenic aerosols.
Figure 2 shows the relationship between CCN 0.4 and CN concentrations.Again, there is a surprisingly good correlation, especially in view of the very different regimes from which the data are taken and the many different instruments by which they have been collected.The data in Table 2 suggest a fairly constant CCN 0.4 /CN ratio of 0.36±0.14(excluding the two very low values from Laramie, Wyoming), which has been plotted as a line in Fig. 2.This relationship reflects the relatively narrow range of hygroscopicity parameters and the convergent character of aerosol size distributions typical of many non-urban regions (Andreae and Rosenfeld, 2008).

Summary and conclusions
Analysis of published and unpublished data on AOT 500 and CCN concentration shows that measurements from remote oceanic and continental regions fall into relatively narrow ranges.Remote marine CCN 0.4 concentrations are typically near or slightly above 100 cm −3 in biologically productive regions and seasons, and of the order of a few tens per cm 3 in winter.The average AOT 500 over the remote oceans is 0.057±0.023,again with lower values in winter.Remote continental areas have, on average, almost twice as many CCN, and a mean AOT 500 of 0.075±0.025(Table 2).Some, or maybe even most of this difference, must be related to the closer proximity that even remote continental sites have to pollution sources, underscoring that the difference between marine and continental levels must have been smaller before the advent of anthropogenic pollution.Support for this statement comes from observations in remote regions that show the presence of residual pollution aerosols even under quite clean conditions, from analyses of the spatial patterns of AOT over continents and oceans, and from global aerosol modeling studies (Andreae, 2007a, b;Andreae and Rosenfeld, 2008).CCN concentrations over polluted regions are on average about one order of magnitude greater than over their remote counterparts, while the AOT 500 values over the polluted regions are about 5 times those over their remote equivalents (Table 2).CCN 0.4 concentrations and AOT 500 values show a surprising degree of correlation, which can be expressed as a power law (note that regions dominated by desert dust have been excluded from this analysis).Given the difficulty of making direct CCN measurements, this relationship should be of great practical value in large-scale studies on the influence of the various direct and indirect aerosol effects on climate, as it provides an easily measured proxy for CCN concentrations.It is obvious, that the use of such a relationship requires a number of caveats, in particular in relation to the statistical nature of the relationship, which includes the considerable variability that must be expected at any given time and place due to the effects of the vertical structure of the aerosol distribution, the influence of relative humidity, the presence of large amounts of mineral dust, and other factors.Various refinements to this analysis can be suggested, including the use of different AOT measurement wavelengths and spaceborne remote sensing.
As noted in the introduction, this study was motivated by the desire to explore the range of values typical of various aerosol regimes in the present-day atmosphere, based on a compilation of observations, with the intention of using the results in modeling studies.The analysis presented here shows that the selected proxies for the microphysical and radiative effects of clouds on aerosols, CCN 0.4 and AOT 500 , do not vary fully independently of one another.This implies that the radiative and microphysical effects of aerosols on clouds, and therefore on climate and precipitation, are correlated and should not be treated independently of one another, at least not on larger scales.Further discussion of this conceptual approach and initial results can be found in the recent paper by Rosenfeld et al. (2008).

Fig. 1 .
Fig.1.Relationship between AOT 500 and CCN 0.4 from investigations where these variables have been measured simultaneously, or where data from nearby sites at comparable times were available.The error bars reflect the variability of measurements within each study (standard deviations or quartiles).

W
. Birmili, A. Wiedensohler, and J. Heintzenberg for letting me use their unpublished CCN and aerosol data.I thank the following AERONET investigators and their staff for establishing and maintaining the sites used in this investigation and for making their data available: B. Holben, W.-M. Hao, B. McArthur, M. Panchenko, P. Goloub, R. Frouin, O. Mayol-Bracero, E. Cuevas, S.Jones, P.-H.Lin, J. E. Nichol, A. Royer, N. O'Neill, S.-C.Yoon.Edited by: W. Conant This Open Access Publication is financed by the Max Planck Society.

Table 2 .
CCN 0.4 , CN, and AOT 500 measurements from remote and polluted, marine and continental environments.Data are presented as means and standard deviations (mmm±sss) or medians and quartile ranges [nnn (lll-uuu)]."AE:" identifies the AERONET site from which the AOT data are taken.Where the CCN and AOT data are from different locations, this is indicated in the Position column.
Scatterplot of CN vs. CCN 0.4 based on the data from Table2.The line represents the mean CCN 0.4 /CN ratio of 0.36.