Abstract

More than 140 supercooled clouds were compared with corresponding out-of-cloud cloud condensation nuclei (CCN) measurements. In spite of significant differences in altitude, temperature, distances from cloud base, updraft velocity (W), entrainment, and so on, the correlation coefficients (R) between droplet and CCN concentrations were substantial although not as high as those obtained in warm clouds with less variability of nonaerosol influences. CCN at slightly lower altitudes than the clouds had higher R values than CCN measured at the same altitude. Ice particle concentrations appeared to reduce droplet concentrations and reduce R between CCN and droplet concentrations, but only above 6-km altitude and for temperatures below −20°C.

Although higher CCN concentrations generally resulted in higher droplet concentrations, increases in droplet concentrations were generally less than the increases in CCN concentrations. This was apparently due to the expected lower cloud supersaturations (S) when CCN concentrations are higher as was usually the case at lower altitudes. Cloud supersaturations showed more variability at higher altitudes and often very high values at higher altitudes. The use of liquid water content rather than droplet concentrations for cloud threshold resulted in higher R between CCN and droplet concentrations. The same R pattern for cumulative droplet–CCN concentrations as a function of threshold droplet sizes as that recently uncovered in warm clouds was found. This showed R changing rapidly from positive values when all cloud droplets were considered to negative values for slightly larger droplet size thresholds. After reaching a maximum negative value at intermediate droplet sizes, R then reversed direction to smaller negative or even positive values for larger cloud droplet size thresholds. This R pattern of CCN concentrations versus cumulative droplet concentrations for increasing size thresholds is consistent with adiabatic model predictions and thus suggests even greater CCN influence on cloud microphysics.

1. Introduction

The indirect aerosol effect (IAE) continues to be the largest climate uncertainty (Alley et al. 2007). It has been well established that cloud condensation nuclei (CCN) concentrations (NCCN) have a large influence on cloud microphysics (e.g., Squires 1956, 1958; Twomey and Warner 1967; Yum and Hudson 2002). Sometimes the aerosol influence is obvious, such as with ship tracks (Hobbs et al. 2000; Hudson et al. 2000), and sometimes aerosol influence appears to be mitigated by cloud dynamics (Baker et al. 1980; Telford and Chai 1980; Telford and Wagner 1981). Since updraft velocity (W) at cloud base interacts with CCN spectra to determine initial cloud droplet concentrations (Nc) (e.g., Twomey 1959, 1977), W variations can mitigate or obscure CCN influence on Nc (e.g., Peng et al. 2005). The tendency for initial Nc to persist, even as condensation continues as air rises beyond cloud base, is because the presence of the initial droplets inhibits nucleation and growth of new droplets, which cannot compete for condensation with established droplets. Thus, most subsequent condensation occurs only on the droplets that were initially produced. This can result in constant Nc with altitude (e.g., Rogers and Yau 1989). However, droplet sizes and concentrations can subsequently be reduced by the evaporation due to entrainment and mixing. Since these dynamic processes are generally independent of the aerosol, they could reduce Nc in a random manner independent of the initial droplet sizes or concentrations. This could eventually reduce the influence of the initial aerosol on Nc (Kim et al. 2008). Furthermore, ice particles are expected to grow at the expense of liquid droplets and this could reduce droplet sizes and thus detected Nc, which could also reduce the influence of the aerosol on Nc.

Therefore, one of the most important and interesting aerosol observations is the comparison of CCN spectral concentrations with initial or adiabatic cloud parcels that have not had their Nc reduced by entrainment (e.g., Yum et al. 1998; Hudson and Yum 2001) or ice particles. When done correctly this can provide a determination of the initial cloud supersaturation S and thus of which CCN (according to S) actually produced the cloud droplets. This is a difficult experiment mainly because only a limited number of clouds or cloud parcels are adiabatic; that is, their Nc is not reduced by entrainment. Moreover, without vertical profiles through cloud base, which are seldom available, it is very difficult to be sure or prove that a given cloud parcel is adiabatic. Therefore, since most clouds (parcels) are not adiabatic, it is more realistic and climatically significant to determine the relative influence of the aerosol (CCN) on clouds at large regardless of their adiabaticity. After all, it is not just the adiabatic parcels of clouds that determine planetary albedo and precipitation. Just because an aerosol influence is less obvious does not diminish its importance. Ideally both adiabatic and nonadiabatic parcels would be sampled and compared with CCN spectra and W.

2. Measurements

We present comparisons of NCCN (Table 1) with Nc and droplet concentrations using various thresholds (Table 2). Regression correlations are shown and concentration comparisons are made. These measurements were all done from the National center for Atmospheric Research (NCAR) C-130 airplane during the Ice in Layer Clouds Experiment (ICE-L). Twelve research flights were done between 7 November and 16 December 2007 over Colorado and Wyoming. Many of the high-altitude measurements were done in Rocky Mountain wave clouds. Although temperatures were always subfreezing (Fig. 1f) there were usually substantial liquid droplets. Cloud droplet measurements presented here were made with the Droplet Measurement Technologies (DMT) cloud droplet probe (CDP) (McFarquhar et al. 2007; Rosenfeld et al. 2008) (range 2.8–47-μm diameter), which has been described as a miniature forward scattering spectrometer probe (FSSP). McFarquhar et al. (2007) found that the CDP concentrations compared well with a CAS probe (2% disagreement) in no ice conditions and seemed to be less susceptible to artifacts when ice was present. Rosenfeld et al. (2008) indicates 1–2-μm CDP size resolution. Liquid water content (LWC) was integrated from the CDP. The CDP operated throughout the final 11 (from 13 November) of the 12 ICE-L research flights. Larger cloud particles (drops or ice) were measured with the 2DC probe (range 25–600-μm diameter; Strapp et al. 2001).

Table 1.

Out-of-cloud average total particle (CN) and cumulative CCN concentrations (cm−3 at altitude) at the various S for all altitudes and for the three altitude bands. Also shown are the total numbers of cloud penetrations and the number of seconds of CCN measurements. The column labeled CCN is for an uncalibrated S greater than 1.5%; K is the slope of the log–log plot of cumulative CCN concentrations vs S over the 0.3%–1% range.

Out-of-cloud average total particle (CN) and cumulative CCN concentrations (cm−3 at altitude) at the various S for all altitudes and for the three altitude bands. Also shown are the total numbers of cloud penetrations and the number of seconds of CCN measurements. The column labeled CCN is for an uncalibrated S greater than 1.5%; K is the slope of the log–log plot of cumulative CCN concentrations vs S over the 0.3%–1% range.
Out-of-cloud average total particle (CN) and cumulative CCN concentrations (cm−3 at altitude) at the various S for all altitudes and for the three altitude bands. Also shown are the total numbers of cloud penetrations and the number of seconds of CCN measurements. The column labeled CCN is for an uncalibrated S greater than 1.5%; K is the slope of the log–log plot of cumulative CCN concentrations vs S over the 0.3%–1% range.
Table 2.

Average cloud microphysics values: number of clouds, number of seconds of cloud measurements, LWC (g m−3), mean diameter (MD, μm) of cloud droplet spectra, and Nc cloud droplet concentrations (cm−3 at altitudes) using droplet concentration 1 cm−3 and various LWC thresholds indicated in square brackets. The clouds, seconds, LWC, and MD are for cloud threshold LWC = 0.01 g m−3.

Average cloud microphysics values: number of clouds, number of seconds of cloud measurements, LWC (g m−3), mean diameter (MD, μm) of cloud droplet spectra, and Nc cloud droplet concentrations (cm−3 at altitudes) using droplet concentration 1 cm−3 and various LWC thresholds indicated in square brackets. The clouds, seconds, LWC, and MD are for cloud threshold LWC = 0.01 g m−3.
Average cloud microphysics values: number of clouds, number of seconds of cloud measurements, LWC (g m−3), mean diameter (MD, μm) of cloud droplet spectra, and Nc cloud droplet concentrations (cm−3 at altitudes) using droplet concentration 1 cm−3 and various LWC thresholds indicated in square brackets. The clouds, seconds, LWC, and MD are for cloud threshold LWC = 0.01 g m−3.
Fig. 1.

Mean values as a function of altitude for each of the 143 clouds that meet the 0.01 g m−3 LWC threshold: (a) Cloud droplet concentrations (Nc); (b) LWC; (c) nearby out-of-cloud CCN concentration at 1% S (N1%); (d) updraft velocity (W) within cloud; (e) CN concentration (NCN) in the same location as the CCN measurements; and (f) cloud temperature.

Fig. 1.

Mean values as a function of altitude for each of the 143 clouds that meet the 0.01 g m−3 LWC threshold: (a) Cloud droplet concentrations (Nc); (b) LWC; (c) nearby out-of-cloud CCN concentration at 1% S (N1%); (d) updraft velocity (W) within cloud; (e) CN concentration (NCN) in the same location as the CCN measurements; and (f) cloud temperature.

CCN were measured with the Desert Research Institute (DRI) CCN spectrometer (Hudson 1989). Total particle concentrations or condensation nuclei (CN) were measured with a TSI3010 condensation nucleus counter. Employment of the CCN spectrometer for special sample processing measurements (Hudson and Da 1996; Hudson 2007) during some of the flights (especially later flights) reduced the available ambient measurements. The cloud data shown are limited to cloud penetrations for which nearby CCN measurements were available within 1–5 min (6–30 km). The CCN measurements that are presented were either at the same altitude as the corresponding cloud measurements or at slightly lower altitudes though not always directly under the corresponding clouds. In some cases two or three different CCN measurement periods (usually before and after the cloud measurement) are averaged. More than 3 h of cloud data (Table 2) in 143 cloud penetrations with an average duration of 70 s and a median duration of 35 s are presented. The CCN measurements associated with each cloud penetration also amounted to more than 3 h (Table 1). These numbers apply for the lowest cloud threshold of 0.01 g m−3 LWC. An LWC threshold of 0.10 g m−3 yielded 2 h of cloud data in 94 penetrations of average duration 73 s and median duration 37 s (Table 3).

Table 3.

Regression coefficients for plots of cloud droplet concentrations against CCN concentrations at 1% S. Each column pertains to average droplet concentrations in each cloud for the seconds with LWC in excess of the values listed in the first row in g m−3. The intercept, slope, and correlation coefficient (R1) pertain to linear regressions of the data. The number of clouds is the number of data points considered; R3 values are correlation coefficients for third-order regressions in droplet concentrations vs CCN. The second row shows the number of cloud penetrations and the third row the total numbers of seconds of those penetrations.

Regression coefficients for plots of cloud droplet concentrations against CCN concentrations at 1% S. Each column pertains to average droplet concentrations in each cloud for the seconds with LWC in excess of the values listed in the first row in g m−3. The intercept, slope, and correlation coefficient (R1) pertain to linear regressions of the data. The number of clouds is the number of data points considered; R3 values are correlation coefficients for third-order regressions in droplet concentrations vs CCN. The second row shows the number of cloud penetrations and the third row the total numbers of seconds of those penetrations.
Regression coefficients for plots of cloud droplet concentrations against CCN concentrations at 1% S. Each column pertains to average droplet concentrations in each cloud for the seconds with LWC in excess of the values listed in the first row in g m−3. The intercept, slope, and correlation coefficient (R1) pertain to linear regressions of the data. The number of clouds is the number of data points considered; R3 values are correlation coefficients for third-order regressions in droplet concentrations vs CCN. The second row shows the number of cloud penetrations and the third row the total numbers of seconds of those penetrations.

Figure 1 shows the vertical distributions of mean values for each cloud penetration. All data considered here are averages of 1-s means. In Figs. 1a, 1b, 1d, and 1f the only 1-s data that is considered for any of these averages is that for which CDP LWC exceeded 0.01 g m−3, which is the cloud threshold. Subsequent presentations with this or other cloud thresholds are done in the same manner. The CN and CCN measurements are averages over periods of approximately 20–400 s outside of clouds. The sample integration time for the CN measurements was one second and it was one to a few seconds for the CCN measurements. The smaller decrease with altitude of Nc (Fig. 1a) compared to the NCCN (Fig. 1c) and CN concentrations (NCN) (Fig. 1e) probably reflects the lower cloud S at lower altitudes due to greater competition among droplets. Thus, at lower altitudes Nc values are lower relative to NCN or NCCN. At lower altitudes where NCCN are higher, Nc are closer to NCCN at lower S. However, some of this difference in vertical gradients might also be due to coincidence in the CDP that might occasionally underestimate high Nc. Another factor is that some of the smaller cloud droplets at high concentrations were below the threshold diameter of the CDP (2.8 μm). Hudson and Yum (2001) showed that even some activated cloud droplets can sometimes be missed by cloud probes. Since most of the clouds that were penetrated in ICE-L were rather small, especially in vertical extent, this means that the droplets were small. Furthermore, the small sizes of the clouds probably made them more susceptible to entrainment and evaporation, which makes the droplets more likely to be small and possibly smaller than the instrument threshold size. The overall average Nc of 119 cm−3 using cloud threshold 0.01 g m−3 LWC (Table 2) was higher when higher LWC thresholds were used (Table 2). Overall average NCCN at 1% S (N1%) was 248 cm−3 (Table 1), which seems to suggest that overall average cloud S was lower than 1% even for an LWC threshold as high as 0.20 g m−3 with average Nc of 190 cm−3. The variability of Figs. 1a, 1c, and 1e makes these comparisons of overall averages of limited value.

A cursory examination of Fig. 1, suggests a convenient altitude division at 3 and 6 km. Tables 1 and 2 show how the average concentrations in these three altitude bands decrease with altitude. Comparing average Nc with average NCCN within each altitude band suggests cloud S less than 0.3% for the lowest altitudes but cloud S exceeding 1.5% at the middle and high altitudes. Average Nc are higher for higher LWC thresholds except for the two highest LWC thresholds in the highest altitude band (Table 2, last columns); this is probably due to the small number of clouds (Table 2, column 1) and the small number of seconds of data that exceed these thresholds (Table 3, row 2). Higher LWC thresholds should better represent cloud parcels that are closer to adiabatic because the higher LWC is probably due to less entrainment, which usually reduces LWC and Nc.

3. Results

Figure 2 compares Nc and N1% for each of the 143 cloud penetrations (LWC > 0.01 g m−3). The linear regression coefficients for Fig. 2 are shown in Table 3, column 1; i.e., the zero N1% intercept of Nc, slope, and correlation coefficient (R). Since these data do not seem to conform to a linear relationship, the R for a third-order polynomial fit of Nc versus N1% is also displayed. Table 3 then also lists these coefficients for plots of Nc versus N1% using higher LWC thresholds. The R values for the second power of Nc versus N1% are not presented because they show unrealistic decreasing Nc for large N1%. The right side of Table 3 shows similar corresponding regression coefficients when data are confined to clouds where NCCN are from slightly lower altitudes than the cloud measurements. The number of clouds and seconds of measurements for each LWC threshold is thus lower. The R values are significantly higher on the right-hand side of Table 3 because the CCN that affect Nc come into the clouds from below.

Fig. 2.

Mean droplet concentrations (Nc) vs nearby out-of-cloud CCN concentrations at 1% supersaturation for all altitudes for the 143 clouds that meet the 0.01 g m−3 LWC threshold. The regression coefficients are in Table 3, column 1. Data are divided into the three altitude bands listed.

Fig. 2.

Mean droplet concentrations (Nc) vs nearby out-of-cloud CCN concentrations at 1% supersaturation for all altitudes for the 143 clouds that meet the 0.01 g m−3 LWC threshold. The regression coefficients are in Table 3, column 1. Data are divided into the three altitude bands listed.

Since there are obvious differences in concentrations with altitude (e.g., higher concentrations in the boundary layer) and since the relationship between Nc and NCCN also varies with concentration (nonlinear), it seems that charting this relationship over such wide ranges may not be of much predictive value. Nonetheless, the relationship indicates the overall importance of NCCN in determining Nc.

Figure 3 shows NcN1% relationships within each of the three altitude bands. Table 4 shows the linear regression coefficients for the three altitude bands using two LWC thresholds both for all clouds and for only those with CCN measurements at slightly lower altitudes. For each of the three altitude bands R is lower than it is when data from all altitudes were considered (Table 3). This is largely because of the smaller range of concentrations within each altitude band. The smaller range is not as obvious for the lowest altitude band where concentrations were never as low as in the other altitude bands. Since there is no suggestion of nonlinearity for the data within each altitude band, higher-order regressions are not considered. The differences between the regressions for all altitudes and for each altitude band demonstrate that measurements over wider concentration ranges tend to produce higher R. This is especially the case when there are so many other variable factors involved in determining Nc such as W, altitude, cloud-base temperature, distances between the cloud measurements and cloud base, and amount of entrainment. Nevertheless, at least R is positive for all but the highest altitude band and lowest LWC threshold. At this altitude the range of variability and number of clouds and seconds is the lowest. Furthermore, even these R values turn positive when the three exceptionally low Nc (Fig. 3c), all from the last flight (RF12, lowest three red squares), are removed from consideration (Table 4, row 4). These three clouds are at the lowest temperatures (upper left corners of Figs. 1f and 1a) of the 143 clouds presented here and they had the highest, second-highest, and fourth-highest 2DC probe concentrations (N2DC larger than 87 μm) of the 21 clouds above 6-km altitude. For the 0.10 g m−3 threshold, the R values in Table 4 are positive for all altitudes because there are no clouds from the last flight that meet this threshold.

Fig. 3.

As in Fig. 2, but for each altitude band: (a) <3 km; (b) 3–6 km; (c) >6 km. In (c) the four red squares are clouds with the highest N2DC > 87 μm diameter. Numbers to the right of each panel are number of clouds.

Fig. 3.

As in Fig. 2, but for each altitude band: (a) <3 km; (b) 3–6 km; (c) >6 km. In (c) the four red squares are clouds with the highest N2DC > 87 μm diameter. Numbers to the right of each panel are number of clouds.

Table 4.

As in Table 3, but for the three altitude bands. The extra row for 0.01 > 6 km excludes RF12.

As in Table 3, but for the three altitude bands. The extra row for 0.01 > 6 km excludes RF12.
As in Table 3, but for the three altitude bands. The extra row for 0.01 > 6 km excludes RF12.

On the left side of Table 4, the lower altitude bands have higher R. When only the below-cloud CCN measurements are considered on the right side of Table 4, R is higher, especially for the mid-altitude band (0.40–0.69 and 0.39–0.69 km) where a large number of the CCN measurements at the same altitudes as the corresponding clouds were apparently not representative of the CCN concentrations that produced the cloud droplets (half of these cases were from RF2). The exception to this is the lowest altitude band and lower LWC threshold where R for below-cloud CCN is lower than R for all CCN.

In mixed phase clouds diffusion of water molecules from droplets to ice particles should produce smaller droplets and thus reductions of droplet concentrations larger than specific sizes. The presence of ice thus could disrupt correlations between NCCN and Nc since ice concentrations may be independent of Nc or NCCN. The 2DC probe measures cloud particles that are larger than most cloud droplets and in much lower concentrations than cloud droplets. At sufficiently cold temperatures many of these particles are ice particles. When the four clouds (red squares in Fig. 3c) with the lowest Nc (<47 cm−3) as well as the highest N2DC larger than 87 μm (>12 L−1 whereas the next lower N2DC in this altitude band had N2DC of only 6 L−1) are removed from consideration, R goes from −0.01 to +0.49 in Fig. 4a. Similarly Table 4, row 4 shows R going to 0.47 when just the three lowest Nc clouds are removed from consideration. When the four clouds with the next higher N2DC are removed from consideration in Fig. 4b, R increases to 0.63. Then when the clouds with the next four higher N2DC are removed, leaving the nine clouds with the lowest N2DC (<0.9 L−1) in Fig. 4c, R goes up to 0.83. This same analysis using a 0.05 g m−3 threshold shows similar results except that only one of the three −35°C clouds meets this criterion. These omissions result in a slight positive R of 0.20 for the 17 clouds that meet this higher LWC threshold. Similar increases of positive R are found by peeling away clouds with next higher N2DC. The 2DC measurements at high altitudes suggest that ice crystals are reducing the droplet concentrations and thus reducing R for N1%Nc. At the two lower altitude bands, equal divisions of the clouds according to N2DC produce similar positive R for N1%Nc that is, R for N1%Nc is not related to N2DC because at these altitudes and temperatures N2DC presumably represents large drops rather than ice particles.

Fig. 4.

As in Fig. 3c, but dividing the 21 clouds according to N2DC > 87 μm: (a) the 17 clouds with the lowest N2DC(<7 L−1) (i.e., Fig. 3c without the 4 red squares); (b) the 13 lowest N2DC (<1.6 L−1); (c) the 9 clouds with the lowest N2DC (<0.9 L−1).

Fig. 4.

As in Fig. 3c, but dividing the 21 clouds according to N2DC > 87 μm: (a) the 17 clouds with the lowest N2DC(<7 L−1) (i.e., Fig. 3c without the 4 red squares); (b) the 13 lowest N2DC (<1.6 L−1); (c) the 9 clouds with the lowest N2DC (<0.9 L−1).

Often droplet number concentration has been used for cloud threshold (e.g., Hudson and Yum 2001). Table 5 shows regressions using 1 cm−3 as the cloud threshold for Nc. This threshold revealed six more clouds and 30% more seconds of data (unequally distributed among the altitudes) than the 0.01 g m−3 threshold. Table 2 shows that average Nc values using the number concentration threshold were approximately 15% lower than Nc using the 0.01 g m−3 LWC threshold. Except for the highest altitude band the R values in Table 5 are significantly lower than corresponding R in Tables 3 and 4, especially for the lowest altitude band. Table 6 shows little improvement of R with higher Nc thresholds. These R are still less than R using even the lowest LWC threshold of 0.01 g m−3 except at the highest altitudes, where none of the R values are very high. Tables 3 and 4 show that higher LWC thresholds result in higher R at least up to 0.10 g m−3, above which the smaller number of clouds and seconds seem to diminish R. These results seem to confirm that LWC thresholds provide a better approach to cloud adiabaticity than Nc thresholds.

Table 5.

Similar to parts of Tables 3 and 4, but for regressions of Nc vs N1% using a cloud threshold of 1 droplet per cm3.

Similar to parts of Tables 3 and 4, but for regressions of Nc vs N1% using a cloud threshold of 1 droplet per cm3.
Similar to parts of Tables 3 and 4, but for regressions of Nc vs N1% using a cloud threshold of 1 droplet per cm3.
Table 6.

Similar to Table 5, but for cloud thresholds of 5 and 10 droplets per cm3.

Similar to Table 5, but for cloud thresholds of 5 and 10 droplets per cm3.
Similar to Table 5, but for cloud thresholds of 5 and 10 droplets per cm3.

Table 7 shows that regressions of Nc (0.01 g m−3 LWC) with NCN have higher intercepts, smaller slopes, and lower R than N1% in Tables 3 and 5. The only exception is the lowest altitude for below cloud CN. Table 7 also shows that NCCN (0.3%) provides slightly lower intercepts and higher slopes and R than N1% for the 0.01 g m−3 LWC threshold.

Table 7.

As in Table 4 for only 0.01 g m−3 LWC threshold, but for CN and CCN at 0.30% S.

As in Table 4 for only 0.01 g m−3 LWC threshold, but for CN and CCN at 0.30% S.
As in Table 4 for only 0.01 g m−3 LWC threshold, but for CN and CCN at 0.30% S.

Comparisons of NCCN spectra (concentrations as a function of S) with Nc reveal the cloud effective S (Seff) (Hudson 1983). In most cases this is an underestimate of the maximum cloud S that produced the initial Nc because some of the droplets may have evaporated below the instrument threshold size. Table 8 shows coarse Seff estimates in ICE-L. As expected at low altitudes where concentrations are higher, Seff values are lower. The middle and high altitudes show a wider range of Seff and higher Seff, mainly because of lower concentrations at the higher altitudes. Note that Seff is higher for the higher Nc at the higher LWC threshold (0.20 compared with 0.01 g m−3). The most puzzling aspect of Table 8 is the large number of clouds in the last column where Nc exceeds even NCN. Half of these for 0.01 g m−3 LWC threshold and eight of the nine for 0.2 g m−3 LWC threshold occurred during RF2. These RF2 anomalies occurred in the mid-altitude range at moderate temperatures (−5° to −11°C) with extremely low or zero N2DC. These cases showed the greatest exceedances of Nc over NCN (up to a factor of 3). For all but one of these clouds the CN and CCN measurements took place at the same altitudes as the clouds. Apparently they were especially irrelevant to the cloud microphysics because they contributed heavily to the low R of 0.40 in Fig. 3b (row 2, Table 4). When most of these cases were ignored by using only CCN measurements from below cloud, R significantly increased to 0.69 (rightmost column of row 2, Table 4). The other exceedances of Nc over NCN were more modest. The four Nc exceedances of NCN above 6 km ranged only up to 27%. For all of these four clouds the CN and CCN measurements were made below cloud, which for these wave clouds would seem to be less relevant to cloud microphysics than measurements at the same altitude. Nonetheless, even though the most anomalous cases were probably due to less than relevant locations of the out-of-cloud aerosol measurements, there were still many cases of rather high Seff, and thus even initial cloud S appears to be so high that virtually all particles became activated cloud droplets. It is unlikely that shattering of ice particles on the CDP produced anomalous high counts because this instrument was specifically designed to minimize this problem and this seemed to be verified by McFarquhar et al. (2007). Moreover, the Nc at high altitudes was not so high or so variable as to suggest very much of a shattering artifact effect on measured Nc. These comparisons also suggest the possibility that the CN counter was not measuring all of the particles, especially at high altitudes where diffusion losses within sample inlet tubes are greater. This is especially so for the smaller CN particles.

Table 8.

Number of clouds with approximate Seff values for two LWC (g m−3) thresholds and for the three altitude bands.

Number of clouds with approximate Seff values for two LWC (g m−3) thresholds and for the three altitude bands.
Number of clouds with approximate Seff values for two LWC (g m−3) thresholds and for the three altitude bands.

Table 9 shows R of Nc with W and the variability of W as expressed by the standard deviation of W (σw), which is a measure of turbulence. These R are shown for two LWC thresholds for the three altitude bands. The most significant R values are at the lowest altitude band (<3 km). The relationships for the two highest R are displayed in Fig. 5. There are high R values also for σw at the highest altitude band for the lower LWC threshold, but these are due only to the extremely low Nc of the last flight, which as shown earlier resulted from the action of ice on the droplet concentrations. When these 2 of 18 and 1 of 9 clouds are removed from consideration in the next row of Table 9, R is negligible. In all cases the R for W with Nc is greater for the higher LWC threshold. The R values for within-cloud W with Nc for each altitude band are similar to the R values for N1% with Nc in Table 4. The number of clouds is smaller in some of the rows of Table 9 compared to corresponding rows of Tables 3 and 4 because of the elimination of 1-s clouds. However, elimination of these clouds had miniscule effects on the R values in Tables 3 and 4. However, the WNc R values for all of the clouds in Table 9 are considerably lower than the corresponding N1%Nc R values in Table 3. This all-cloud contrast in R is due to the aforementioned greater range of N1% when all clouds are considered. In other words, there are greater differences in CCN concentrations at the various altitudes whereas the W values are more similar at the various altitudes. Thus, it is inappropriate to try to correlate W or σw with Nc over wide differences in altitude or NCCN or Nc.

Table 9.

As in Table 4, but the regressions are of mean updrafts (W) and standard deviations of W (σw) with droplet concentrations (Nc). The below cloud category represents W measurements made when the CCN were measured only below adjacent clouds but not necessarily immediately below each cloud.

As in Table 4, but the regressions are of mean updrafts (W) and standard deviations of W (σw) with droplet concentrations (Nc). The below cloud category represents W measurements made when the CCN were measured only below adjacent clouds but not necessarily immediately below each cloud.
As in Table 4, but the regressions are of mean updrafts (W) and standard deviations of W (σw) with droplet concentrations (Nc). The below cloud category represents W measurements made when the CCN were measured only below adjacent clouds but not necessarily immediately below each cloud.
Fig. 5.

As in Fig. 3a (altitude < 3 km), but for LWC > 0.1 g m−3: (a) Nc is plotted against updraft velocity (W) measured within these 24 clouds; (b) as in (a), but only the 17 clouds with CCN measurements below the clouds are plotted against the standard deviations of W(σw) measured at the same place below the clouds.

Fig. 5.

As in Fig. 3a (altitude < 3 km), but for LWC > 0.1 g m−3: (a) Nc is plotted against updraft velocity (W) measured within these 24 clouds; (b) as in (a), but only the 17 clouds with CCN measurements below the clouds are plotted against the standard deviations of W(σw) measured at the same place below the clouds.

The location of the W measurements was not ideal but it is difficult to measure W at the time and location when and where clouds form and Nc is determined. Here we consider W measurements within cloud or at the location of the below-cloud CCN measurements as surrogates for the W that produced the clouds. We are suggesting that the measured in-cloud or below-cloud W values are proportional to the W that originally produced the clouds. The W and σw within clouds are only for those seconds when LWC exceeds the given values. These had higher W, which is probably more relevant to or more representative of W at cloud formation. Median W and the percentage of clouds with positive average W progressively increased for higher LWC thresholds. Although mean or median W values were often negative, high Nc values were mainly confined to clouds with positive mean W.

Although negative W does not produce clouds, W is not only an indicator of the dynamic contribution to the initial Nc but also an indication of entrainment (i.e., positive W suggests less entrainment that could reduce Nc). Ignoring the negative W values would bias the data. When we considered only clouds with positive average W we eliminated half of the clouds and found no difference in R for N1%Nc or WNc. When we limited each second of data to positive W we retained 90% of the clouds but 50% of the number of seconds and again found only very modest changes in R. Attempts to use the product of W and NCCN (Hudson and Noble 2009; hereafter HN9) resulted in negligible increases in R over those for N1%Nc. The product of σw and NCCN resulted in lower R than R for N1%Nc.

Hudson et al. (2009, hereafter H9) and HN9 showed how R between N1% and cumulative droplet concentrations larger than specific droplet size thresholds (Nt) varies with the threshold sizes. The value of R was positive for small size thresholds that included all activated cloud droplets (Nc) because in this cumulative cloud droplet size range the Nt values are directly related to the concentrations of the nuclei upon which the droplets formed. These are usually NCCN at high S (e.g., N1%) but even for clouds formed at lower S the NCCN(S) are often in proportion to the NCCN at higher S (i.e., N1%). Therefore, R for NtNCCN at most S (i.e., N1%) is positive. At larger droplet size thresholds R was negative (H9 and HN9). The negative R for NtNCCN was the result of competition among droplets for condensed water, which restricts cloud S and droplet sizes to a greater extent when concentrations are higher. The greater restriction of cloud S and droplet sizes in cases when NCCN is higher produces the negative R for Nt with NCCN for Nt above somewhat larger size thresholds. Although for each CCN spectrum NCCN is always lower at lower S, Nc is always higher for higher NCCN situations compared to lower NCCN situations even though the latter produce higher cloud S; this is the case when all other factors are equal.

Droplet sizes are restricted to the greatest extents within and near the size range of the mode of the droplet distribution. This is the phenomenon that turns R for NtN1% from positive when all cloud droplets are considered (i.e., NcNCCN) to negative R for NtN1% for increasing size thresholds; R thus reaches its maximum negative absolute value (HN9) slightly beyond the average mode of the droplet distributions (Fig. 2 of HN9). Competition for condensate is less intense for droplets beyond the mode. The smaller droplets within the mode, though numerous, usually do not have sufficient surface area to have much influence on the growth of the larger droplets. Thus, at sizes larger than the mode of the droplet distributions the concentrations of droplets tend toward being proportional to the concentrations of particles upon which they condensed (i.e., the larger and more soluble particles, CCN with low Sc). Often NCCN at various S are in proportion for the various aerosol distributions. If and when this is the case the R for N1% with the concentrations of large cloud droplets will tend toward less negative values (H9) or even positive values (HN9).

Within cumulative cloud droplet size ranges that are intermediate between the average droplet mode and both the small and large sizes there will be conflict between the positive and negative R tendencies that will result in intermediate R values. The depth and breadth of the negative R depends on the degree of competition, the range of concentration variability, cloud adiabaticity, dynamics, temperature, distance from cloud base, entrainment, and CCN spectra, among others. The sequence of R values that has been described would best be exhibited by cloud measurements at similar distances above cloud bases that are at similar temperatures. That way the droplet spectra in the various aerosol regimes would be at similar stages of development. This was apparently the case for the Rain in Cumulus over the Ocean (RICO) experiment (H9) where cloud-base temperatures and altitudes were similar throughout the project and cloud measurements were compared within similar altitude bands. This was apparently also much the case for the Pacific Atmospheric Sulfur Experiment (PASE; HN9) wherein only very shallow clouds were observed. This is certainly not the case in ICE-L where clouds are considered over large altitude and temperature ranges with no discrimination of distances between cloud base and the cloud measurements. As a result, it seemed less likely to discover negative R between Nc and NCCN in ICE-L. This was especially so when all altitudes were considered together. Nonetheless, even when all altitudes were considered together R displayed the characteristic progressive decrease with increasing droplet size thresholds (Fig. 6), similar to the findings of H9 and HN9. But the negative R in ICE-L has smaller absolute values. However, the number of data points (94) in ICE-L is much larger since they are for individual clouds rather than the smaller number of flight averages in RICO or PASE. There was much more concentration variability within flights in ICE-L than there was in RICO or PASE. Therefore, the significance level of these smaller negative R values approaches 90% though the coefficient of determination (R2) suggests modest influence of NCCN on Nt variations. But even with the large variations of other factors that influence Nc and Nt the effect of the aerosol comes through even for these small negative R values.

Fig. 6.

Correlation coefficients (R) of Nc with N1% for various droplet size thresholds for Nc. This is for clouds at all altitudes using an LWC threshold of 0.10 g m−3. Thus, 94 clouds (data points) are considered for each correlation (Table 3, column 3).

Fig. 6.

Correlation coefficients (R) of Nc with N1% for various droplet size thresholds for Nc. This is for clouds at all altitudes using an LWC threshold of 0.10 g m−3. Thus, 94 clouds (data points) are considered for each correlation (Table 3, column 3).

Figure 7a for the 50 clouds within the intermediate altitude band and LWC >0.10 g m−3 yields a maximum negative R of 0.5, which has a significance level beyond 99.95%. As in the previous studies, this greatest negative R occurs at sizes just beyond the mode of the average spectra shown in Fig. 7b. As in RICO the negative R of N1%Nt in ICE-L is slightly greater than the positive R for N1%Nc; in ICE-L similar positive R values occur at both ends of the droplet size distribution as was the case in PASE. This suggests that there is proportionality— among the NCCNs at various Ss among the input aerosols—to the various cloud parcels considered here in ICE-L, as was indicated in PASE (HN9). The fact that R smoothly changes with size threshold indicates the validity of the correlations.

Fig. 7.

(a) As in Fig. 6, but only the altitude range of 3–6 km; this means 50 clouds (data points) for each correlation (Table 4, column3). (b) Mean differential cloud droplet spectra of the 50 clouds also using the 0.10 g m−3 LWC threshold.

Fig. 7.

(a) As in Fig. 6, but only the altitude range of 3–6 km; this means 50 clouds (data points) for each correlation (Table 4, column3). (b) Mean differential cloud droplet spectra of the 50 clouds also using the 0.10 g m−3 LWC threshold.

Figure 8 demonstrates that the R patterns displayed in Figs. 6 and 7a and in HN9 and H9 are consistent with theoretical predictions of droplet spectra. Here the Robinson (1984) adiabatic droplet growth model is applied to an observed ICE-L CCN spectrum and two spectra that are multiples of that spectrum (i.e., concentrations at all S are in the same proportions). All other factors (altitude, W, pressure, and temperature) are identical for these three predictions. At threshold sizes below 7 μm Nc is proportional to NCCN, between 9 and 12 μm Nc is inversely related to NCCN, and beyond 14 μm Nc is again positively correlated with NCCN.

Fig. 8.

Computer model–predicted droplet concentrations for CCN spectra that are multiples of each other.

Fig. 8.

Computer model–predicted droplet concentrations for CCN spectra that are multiples of each other.

4. Discussion

The NcNCCN R values at the highest altitude band were very low, mainly because of the influence of ice particles on droplet concentrations. The extremely low Nc of less than 10 cm−3 with temperatures less than −35°C may not have been liquid droplets because this is cold enough to produce homogeneous ice (Sassen and Dodd 1988; Heymsfield and Miloshevich 1993). It is rather certain that the N2DC in these three cold clouds were ice particles, especially since these were the highest N2DC of the high-altitude clouds. The low Nc in these clouds may have been liquid or solid: possibly frozen droplets. These low Ncs relative to the nearby out-of-cloud NCNs and NCCNs are commensurate with droplet concentration reductions due to the presence of ice particles.

When the wider range of concentrations at all altitudes were considered in Fig. 2 the NcNCCN regression appeared to be nonlinear; Nc did not keep pace with NCCN at high concentrations. This rolling off of Nc at high NCCN observed previously by Leaitch et al. (1986) and Hudson and Yum (2002) is expected because of the competition among droplets that drives down cloud S when NCCN is higher so that Nc in those cases is proportional to NCCN at lower S, which for the same aerosol is lower than NCCN at higher S. However, as noted earlier these Nc are still higher than the Nc for lower NCCN situations. This was consistent with the comparisons of CCN spectral concentrations with Nc (Table 8). These comparisons indicated considerably lower Seff at lower altitudes where concentrations were higher. At higher altitudes there was considerably more variability of Seff and significantly higher Seff. In fact the Seff was so high in a large proportion of clouds at higher altitudes that every atmospheric particle seemed to act as a nucleus for cloud droplets. This is especially pertinent since Seff is an underestimate of initial maximum cloud S that produced adiabatic Nc. The variability of Seff contributed to the lack of correlation between Nc and NCCN at a fixed S. These observations seem to mitigate the concern that entrainment and/or ice formation had reduced Nc significantly below adiabatic values. The latter was the case for only a few high-altitude clouds with high N2DC.

Correlations of Nc with within-cloud W were lower but usually comparable with R for NcNCCN for the three altitude bands. Also, R for below-cloud W with Nc was considerably lower than the corresponding R values for NCCNNc. Correlations of Nc with σw were good only for the lowest altitude band (<3 km), especially for σw measured below the corresponding clouds even though these were not directly below the clouds. Measurement limitations may have reduced the apparent influence of W on Nc; however, accuracy of W for the NCAR C-130 system is purported to be 0.10 m s−1 (Lenschow and Spyers-Duran 1989) and since we are dealing only in terms of relative values, precision should be closer to 1 cm s−1.

The fact that the LWC threshold rather than droplet concentration threshold provided superior NcNCCN correlations (Tables 3 –6) is somewhat contrary to RICO (H9), where the 1 cm−3 threshold produced R values similar to the LWC thresholds. The similarities of R for NtNCCN for the different LWC thresholds in Table 3 are similar to those in H9. As in RICO and PASE this suggests that entrainment did not disrupt the effect of CCN on cloud microphysics. This is all the more significant for ICE-L where there was so much more variability in the other factors that influence Nc.

Correlations using NCCN at a lower S of 0.3% were only slightly better than the N1% correlations, even at low altitudes where Seff was considerably lower, and a majority of cases in Table 8 indicated Seff <0.3%. This indicated that indeed Seff was an underestimate of the cloud S that had produced the original Nc and thus that Nc at low altitudes was reduced by entrainment. On the other hand, the R similarities of Nc with NCCN at various S may merely reflect the proportionalities of the CCN spectra. These results, nevertheless, call for CCN measurements at lower S at the lower altitudes, which might show better correlations with Nc.

The NtNCCN R pattern was consistent with predictions of droplet spectra for various CCN spectra with the same shape. However, we confirmed that differences in CCN spectral shapes also produce differences in droplet spectral shapes that produce consequent differences in the pattern of R for NtNCCN with threshold size. The fact that observed R patterns are similar to these predicted R patterns suggests that the CCN spectral shapes were not different enough to disrupt the predicted R pattern. Further investigations of measured droplet and CCN spectra and model predictions are warranted because the results suggest greater influences of CCN on droplet spectra than just determination of total cloud droplet concentrations—that is, spectral influences (i.e., Takahashi and Lee 1978; Takeda and Kuba 1982). Furthermore, as pointed out by HN9, the lower R values at intermediate droplet size thresholds do not necessarily indicate less aerosol influence but are the result of the conflict between the positive and negative influences on R (i.e., the direct correlation with aerosol concentrations and the negative R due to competition among droplets). The negative values were not as deep in ICE-L as was the case with the more similar cloud parcels considered in RICO and PASE.

Although there have been several previous CCN measurements in supercooled conditions (Hoppel et al. 1973; Radke et al. 1984; Hudson and Xie 1998; Yum and Hudson 2001) and many cloud microphysics measurements in supercooled conditions, there has only been one previous study that has included both together (Rosenfeld et al. 2008). The present study is more extensive and includes a wider range of concentrations and temperatures. CCN concentrations in ICE-L show altitude differences similar to those observed by Hudson and Xie (1998) except that the decrease with altitude seems more gradual in ICE-L, which may be a function of the different seasons: spring versus fall/winter. The high-altitude N1% concentrations in the two projects were generally similar except that the earlier study found more variability with concentrations of several hundred and less than 10 cm−3. These CCN results in ICE-L indicate that further analysis is justified. This will focus on lower S CCN measurements and attempt to isolate more similar and comparable cloud parcels.

5. Conclusions

In spite of large differences in altitudes, temperatures, dynamics, adiabaticity, entrainment, etc., total cloud droplet concentrations (Nc) in a variety of supercooled clouds at various altitudes over Colorado and Wyoming were correlated with nearby out-of-cloud CCN concentrations (NCCN). Correlations of Nc with NCCN were significantly higher when restricted to clouds with CCN measurements at slightly lower altitudes, since the major CCN influence comes about from air coming into cloud base. Correlation coefficients (R) were stronger for the larger concentration ranges encompassed by all altitudes considered together compared to the smaller ranges of concentrations within each of three altitude bands. However, the ICE-L NCCNNc Rs were not as high as those obtained in warm clouds where there were smaller variations in other factors that influenced droplet concentrations.

High-altitude clouds with high ice concentrations had either negative or negligible NCCNNc Rs; whereas high-altitude clouds with lower ice concentrations had progressively higher positive R. Thus, clouds at high altitudes showed droplet concentration reductions associated with the presence of ice particles.

Correlations between measured updraft velocities (W) and Nc were comparable with NCCNNc Rs, especially at the lowest altitudes. Cloud supersaturations seemed to follow the expected pattern of being higher when CCN concentrations were lower. Liquid water content (LWC) thresholds produced better correlations between CCN and droplet concentrations than droplet concentration thresholds. Higher LWC thresholds, which should correspond to more adiabatic (less entrained) clouds showed higher CCN–droplet correlations.

CCN correlations with cumulative cloud droplet concentrations changed from positive to negative and back to positive with increasing droplet threshold sizes. The negative correlations at intermediate threshold sizes are due to competition among cloud droplets for condensed water, which causes greater limitations to droplet sizes when concentrations are higher. The positive correlations for large size thresholds are attributed to proportional relationships among CCN concentrations at various S and to the lower concentrations of large cloud drops that result in less competition for condensate. This correlation pattern is consistent with predictions of an adiabatic model for various NCCN spectra that have identical shapes. However, since the shapes of CCN spectra are not always identical, the shapes of subsequent droplet spectra may consequently differ enough to cause different patterns of correlations between CCN and cumulative droplet concentrations.

Acknowledgments

This work was supported by NSF Grant ATM-0615414. The National Center for Atmospheric Research provided the C-130 airplane and the CDP, 260X, and CN measurements. Discussions with David Mitchell of DRI were very helpful.

REFERENCES

REFERENCES
Alley
,
R. B.
, and
Coauthors
,
2007
:
Summary for policymakers.
Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 1–18
.
Baker
,
M. G.
,
R. G.
Corbin
, and
J.
Latham
,
1980
:
The influence of entrainment on the evolution of cloud droplet spectra. I. A model of inhomogeneous mixing.
Quart. J. Roy. Meteor. Soc.
,
106
,
581
598
.
Heymsfield
,
A. J.
, and
L. M.
Miloshevich
,
1993
:
Homogeneous ice nucleation and supercooled liquid water in orographic wave clouds.
J. Atmos. Sci.
,
50
,
2335
2353
.
Hobbs
,
P. V.
, and
Coauthors
,
2000
:
Emissions from ships with respect to their effects on clouds.
J. Atmos. Sci.
,
57
,
2570
2590
.
Hoppel
,
W. A.
,
J. E.
Dinger
, and
R. E.
Ruskin
,
1973
:
Vertical profiles of CCN at various geographical locations.
J. Atmos. Sci.
,
30
,
1410
1420
.
Hudson
,
J. G.
,
1983
:
Effects of CCN concentrations on stratus clouds.
J. Atmos. Sci.
,
40
,
480
486
.
Hudson
,
J. G.
,
1989
:
An instantaneous CCN spectrometer.
J. Atmos. Oceanic Technol.
,
6
,
1055
1065
.
Hudson
,
J. G.
,
2007
:
Variability of the relationship between particle size and cloud-nucleating ability.
Geophys. Res. Lett.
,
34
,
L08801
.
doi:10.1029/2006GL028850
.
Hudson
,
J. G.
, and
X.
Da
,
1996
:
Volatility and size of cloud condensation nuclei.
J. Geophys. Res.
,
101
,
4435
4442
.
Hudson
,
J. G.
, and
Y.
Xie
,
1998
:
Cloud condensation nuclei measurements in the high troposphere and in jet aircraft exhaust.
Geophys. Res. Lett.
,
25
,
1395
1398
.
Hudson
,
J. G.
, and
S. S.
Yum
,
2001
:
Maritime-continental drizzle contrasts in small cumuli.
J. Atmos. Sci.
,
58
,
915
926
.
Hudson
,
J. G.
, and
S. S.
Yum
,
2002
:
Cloud condensation nuclei spectra and polluted and clean clouds over the Indian Ocean.
J. Geophys. Res.
,
107
,
8022
.
doi:10.1029/2001JD000829
.
Hudson
,
J. G.
, and
S.
Noble
,
2009
:
CCN and cloud droplet concentrations at a remote ocean site.
Geophys. Res. Lett.
,
36
,
L13812
.
doi:10.1029/2009GL038465
.
Hudson
,
J. G.
,
T. J.
Garrett
,
P. V.
Hobbs
,
S. R.
Strader
,
Y. X.
Xie
, and
S. S.
Yum
,
2000
:
Cloud condensation nuclei and ship tracks.
J. Atmos. Sci.
,
57
,
2696
2706
.
Hudson
,
J. G.
,
S.
Noble
,
V.
Jha
, and
S.
Mishra
,
2009
:
Correlations of small cumuli droplet and drizzle drop concentrations with cloud condensation nuclei concentrations.
J. Geophys. Res.
,
114
,
D05201
.
doi:10.1029/2008JD010581
.
Kim
,
B-G.
,
M. A.
Miller
,
S. E.
Schwartz
,
Y.
Liu
, and
Q.
Min
,
2008
:
The role of adiabaticity in the aerosol first indirect effect.
J. Geophys. Res.
,
113
,
D05210
.
doi:10.1029/2007JD008961
.
Leaitch
,
W. R.
,
J. W.
Strapp
,
G. A.
Isaac
, and
J. G.
Hudson
,
1986
:
Cloud droplet nucleation and scavenging of aerosol sulphate in polluted atmospheres.
Tellus
,
388
,
328
344
.
Lenschow
,
D. H.
, and
P.
Spyers-Duran
,
cited
.
1989
:
Measurement techniques: Air motion sensing.
NCAR RAF Bull. 23. [Available online at http://www.eol.ucar.edu/raf/Bulletins/bulletin23.html]
.
McFarquhar
,
G. M.
,
J.
Um
,
M.
Freer
,
D.
Baumgardner
,
G. L.
Kok
, and
G.
Mace
,
2007
:
Importance of small ice crystals to cirrus properties: Observations from the Tropical Warm Pool International Cloud Experiment (TWP-ICE).
Geophys. Res. Lett.
,
34
,
L13803
.
doi:10.1029/2007GL029865
.
Peng
,
Y.
,
U.
Lohmann
, and
W. R.
Leaitch
,
2005
:
Importance of vertical velocity variations in the cloud droplet nucleation process of marine stratus clouds.
J. Geophys. Res.
,
D21213
.
doi:10.1029/2004JD004922
.
Radke
,
L. F.
,
J. H.
Lyons
,
D. A.
Hegg
,
P. V.
Hobbs
, and
I. H.
Bailey
,
1984
:
Airborne observations of arctic aerosols. I: Characteristics of Arctic haze.
Geophys. Res. Lett.
,
11
,
393
396
.
Robinson
,
N. F.
,
1984
:
The efficient numerical calculation of condensational cloud drop growth.
J. Atmos. Sci.
,
41
,
697
700
.
Rogers
,
R. R.
, and
M. K.
Yau
,
1989
:
A Short Course in Cloud Physics.
Pergamon Press, 293 pp
.
Rosenfeld
,
D.
,
W. L.
Woodley
,
D.
Axisa
,
E.
Freud
,
J. G.
Hudson
, and
A.
Givati
,
2008
:
Aircraft measurements of the impacts of pollution aerosols on clouds and precipitation over the Sierra Nevada.
J. Geophys. Res.
,
113
,
D15203
.
doi:10.1029/2007JD009544
.
Sassen
,
K.
, and
G. C.
Dodd
,
1988
:
Homogeneous nucleation rate for highly supercooled cirrus cloud droplets.
J. Atmos. Sci.
,
45
,
1357
1369
.
Squires
,
P.
,
1956
:
The microstructure of cumuli in maritime and continental air.
Tellus
,
8
,
443
444
.
Squires
,
P.
,
1958
:
The microstructure and colloidal stability of warm clouds. Part II: The causes of the variations in microstructure.
Tellus
,
10
,
262
271
.
Strapp
,
J. W.
,
F.
Albers
,
A.
Reuter
,
A. V.
Korolev
,
U.
Maixner
,
E.
Rashke
, and
Z.
Vukovic
,
2001
:
Laboratory measurements of the response of a PMS OAP-2DC.
J. Atmos. Sci.
,
18
,
1150
1170
.
Takahashi
,
T.
, and
S-M.
Lee
,
1978
:
The nuclei mass range most efficient for the initiation of warm cloud showers.
J. Atmos. Sci.
,
35
,
1934
1946
.
Takeda
,
T.
, and
N.
Kuba
,
1982
:
Numerical study of the effect of CCN on the size distribution of cloud droplets. Part I. Cloud droplets in the stage of condensation growth.
J. Meteor. Soc. Japan
,
60
,
978
992
.
Telford
,
J. W.
, and
S. K.
Chai
,
1980
:
A new aspect of condensation theory.
Pure Appl. Geophys.
,
118
,
720
742
.
Telford
,
J. W.
, and
P. B.
Wagner
,
1981
:
Observations of condensation growth determined by entity type mixing.
Pure Appl. Geophys.
,
119
,
934
965
.
Twomey
,
S.
,
1959
:
The supersaturation in natural clouds and the variation of cloud droplet concentration.
Geophys. Pure Appl.
,
43
,
243
249
.
Twomey
,
S.
,
1977
:
Atmospheric Aerosols.
Elsevier, 302 pp
.
Twomey
,
S.
, and
J.
Warner
,
1967
:
Comparison of measurements of cloud droplets and cloud nuclei.
J. Atmos. Sci.
,
24
,
702
703
.
Yum
,
S. S.
, and
J. G.
Hudson
,
2001
:
Vertical distributions of cloud condensation nuclei spectra over the springtime Arctic Ocean.
J. Geophys. Res.
,
106
,
15045
15052
.
Yum
,
S. S.
, and
J. G.
Hudson
,
2002
:
Maritime/continental microphysical contrasts in stratus.
Tellus
,
54B
,
61
73
.
Yum
,
S. S.
,
J. G.
Hudson
, and
Y.
Xie
,
1998
:
Comparisons of cloud microphysics with cloud condensation nuclei spectra over the summertime Southern Ocean.
J. Geophys. Res.
,
103
,
16625
16636
.

Footnotes

Corresponding author address: James G. Hudson, Division of Atmospheric Science, DRI, 2215 Raggio Pkwy., Reno, NV 89512–1095. Email: hudson@dri.edu