[1] Giant nuclei (GN) concentrations (N_{GN}) below RICO small cumulus clouds were substantially correlated with drizzle drop concentrations (N_{d}), especially at higher cloud altitudes. The N_{GN}-N_{d} correlation coefficients (R) progressively increased with altitude whereas R for CCN concentrations with N_{d} were negative with mostly decreasing magnitudes at increasing altitudes. These results indicate that the positive influence of GN [or CCN with low critical supersaturations (S_{c})] on N_{d} is greater than the negative influence of high S_{c} CCN on N_{d} at high cloud altitudes where there are more drizzle drops. This work has implications not only for fundamental cloud physics but also for climate change; i.e., global warming and the indirect aerosol effect as well as geoengineering and hygroscopic cloud seeding.

[2] Nonfreezing (warm) clouds rain more easily in maritime than continental air [Battan and Braham, 1956]. The lower submicrometer cloud condensation nuclei (CCN) concentrations (N_{CCN}) of maritime air produce fewer and thus larger droplets that are more likely to precipitate [Hudson and Yum, 2001]. However, maritime precipitation is often so rapid that it seems necessary to invoke further explanations [Mason, 1971] involving either dynamics [e.g., Telford et al., 1984; Pinsky and Khain, 2002] or giant nuclei (GN > 1 μm) [e.g., Johnson, 1993; Blyth et al., 2003], which more directly produce precipitable drops. Although GN are known to be produced by the action of wind at the ocean surface [Woodcock and Blanchard, 1955] their role in warm rain has been debated for decades though many recent publications have been negative [e.g., Hudson and Yum, 2001; Göke et al., 2007; Hudson and Mishra, 2007 (hereafter HM7); Knight et al., 2008]. The GN-warm rain hypothesis has implications beyond cloud physics. The most uncertain aspect of the indirect aerosol effect (IAE), 2nd IAE (precipitation inhibition by anthropogenic N_{CCN}) would be further complicated by precipitation enhancement by increased concentrations of GN (N_{GN}). Schickedanz [1974] noted that urban areas are a source of GN while Mather [1991] demonstrated positive influence of anthropogenic GN on warm rain. Latham and Smith [1990] showed that global warming could enhance maritime wind velocities, which could perturb precipitation by increasing N_{GN}.

2. Background

[3] During the Rain in Cumulus over the Ocean (RICO) project [Rauber et al., 2007] N_{CCN} were highly correlated with total cloud droplet concentrations (N_{c}; larger than 2.4 μm diameter) at altitudes up to 3 km (Figure 1a) [Hudson et al., 2009] (hereafter H9). Precipitation inhibition of higher N_{CCN} was demonstrated by negative correlations between N_{CCN} active at a supersaturation (S) of 1% (N_{1%}) with cumulative concentrations of cloud droplets larger than specified threshold sizes (N_{t}) (here diameter > 20 μm in Figure 1a, except 3–5 km band). Moreover, the maxima of these negative linear correlation coefficients (R) were even greater at higher altitudes and occurred at larger cumulative threshold sizes at higher altitudes (Figure 1a, except 3–5 km). These extensions of negative R at higher altitudes are due to the typical overall larger drop sizes at higher altitudes. Negative R for N_{1%}-N_{t} extended to drizzle sizes (diameter > 50 μm, N_{d}), especially at higher altitudes where there was more drizzle (greater N_{d}, Table 1, columns 8–10). However, the magnitudes of these negative R values decreased with larger drizzle drop diameters (Figure 1a).

Table 1. Altitude, Number of Flights, Flight Numbers, Correlation Coefficients (R) Between N_{1%} and N_{GN}, Mean CCN, GN, Total Droplet, and Drizzle Drop Concentrations (N_{d}) in Each Altitude Band^{a}

Altitude (m)

No

Flights

R

N_{1%}

N_{GN}

N_{c}

>45 μm

>205 μm

>295 μm

ΔR

a

Last column shows the difference between R for N_{GN} and N_{1%} with N_{d} larger than 245 μm. Last 3 rows consider only those flights with a monotonic relationship between N_{GN} and N_{1%}.within 3 altitude bands. N_{1%}, N_{GN} and N_{c} units are cm^{−3}; drizzle drop concentration units are per liter.

600–900

15

1,3–15

−0.010

108

0.15

87

6

0.5

0.33

−0.14

900–1200

13

1,3–12,15,18

−0.278

104

0.13

73

6

0.5

0.23

0.37

1200–1500

12

1,3–5,7–12,14,18

−0.024

105

0.14

78

12

0.5

0.20

0.49

1500–1800

11

1, 3, 5, 7, 8, 10, 12–15,18

0.048

112

0.15

79

23

0.8

0.32

0.69

1800–2400

10

1,3,5,7,8,10,11, 13,14,18

−0.193

112

0.15

88

98

2.5

0.83

0.82

2400–3000

6

1,3,7,11,13,15

−0.151

111

0.15

91

206

5.5

2.08

0.87

3000–5000

5

3, 5, 13–15

0.996

114

0.18

84

219

10.1

4.13

0.04

2400–3000 M

4

3, 7, 13, 15

0.998

95

0.15

78

251

6.7

2.60

0.01

1500–1800 M

5

3, 5, 13–15

0.996

114

0.18

68

28

1.0

0.44

0.04

600–900 M

5

3, 5, 13–15

0.996

114

0.18

90

7

0.5

0.37

−0.05

[4]Hudson and Noble [2009] (hereafter HN9) and Hudson et al. [2010] (hereafter H10) also observed similar N_{1%}-N_{t} R patterns in other cloud systems. Negative R for N_{1%}-N_{t} was explained by greater competition among droplets for condensed water when N_{1%} were higher. Higher N_{c} due to higher N_{1%} produced greater limitations to droplet sizes when there was more competition. Thus, the smaller droplets of higher N_{c} situations meant that there were fewer droplets exceeding various larger size thresholds. The maximum negative R of N_{1%}-N_{t} occurred at threshold sizes slightly larger than the mode of the average droplet spectrum (HN9; H10). These negative R for N_{1%}-N_{t} sharply contrast with the positive R for N_{1%}-N_{c} (left side of Figure 1a). However, lower N_{t} for even larger threshold drop sizes; i.e., threshold sizes well beyond the mode of the average droplet spectrum produced less competition for condensate among these sparse drop concentrations. This resulted in a smaller negative R tendency for these N_{t} with N_{1%}. Therefore, R for these lower N_{t} at these much larger threshold sizes reverted back toward positive with N_{1%} because, like N_{c}, the sparse concentrations of the very large drops tend to be in proportion to the concentrations of the nuclei upon which they had condensed. The nuclei of these very large drops, however, should be a smaller CCN subset, only CCN with lower critical supersaturations (S_{c}) than the CCN that were responsible for N_{c} (all drops); i.e., N_{CCN} at high S; i.e., 1% (N_{1%}). GN are the lowest S_{c} CCN with the smallest concentrations. N_{t} at the largest threshold sizes should be proportional to N_{GN}.

[5] When N_{GN} are proportional to N_{1%} there should also be a positive R tendency for N_{1%} with the low N_{t} at very large drop size thresholds. These explanations are then consistent with observations of positive R for N_{1%} with N_{c} (i.e., small droplet size thresholds) and with N_{t} at large cumulative drop sizes where there is little competition for condensate; and negative R for N_{1%}-N_{t} at intermediate size thresholds where there is the largest effect of competition among droplets for condensate. H10 demonstrated adiabatic droplet growth model predictions that reproduced these observed R patterns when similar CCN spectral shapes were assumed for each of the various cloud situations that were considered. This means N_{CCN} in the same proportions for all S for the entire data set under consideration. The extreme of N_{CCN} proportionality is N_{GN} ∝ N_{1%}.

[6]Colón-Robles et al. [2006] (hereafter CR6) examined the influence of GN on RICO cloud droplet spectra (diameter < 50 μm) within the lowest cloud altitude band (<900m) and determined that the concentrations of largest cloud droplets were opposite of that expected for a GN influence. Now H9 analysis is extended above 3 km and the GN measurements of CR6 are included. As in most of H9, cloud is defined as liquid water content (LWC > 0.1 gm^{−3}).

3. Results of Analysis

[7]Figure 1 shows R for N_{1%} (panel A) and N_{GN} (panel B) with N_{t} within seven altitude bands. The uniformity of the below cloud aerosol during each RICO flight allowed consideration of flight averages of N_{1%}, N_{GN}, N_{c} and N_{t} [HM7 and H9]. N_{GN} are averages of the morning and afternoon concentrations in Figure 1a of CR6. Another flight is added by using GN measurements of Reiche and Lasher-Trapp [2010]. The R of 0.97 between these N_{GN} and N_{GN} of CR6 for the 9 flights with common data made this possible.

[8] The 3–5 km altitude band shows a different R pattern for N_{d} from that of the lower altitudes (right hand side of Figure 1a). This apparent positive influence of N_{1%} on N_{d} is inconsistent with the negative influence of N_{1%} on large threshold N_{t} observed and described by H9, HN9 and H10. The positive R for N_{GN} with N_{d} at the highest altitude band in Figure 1b is consistent with the explanations of the last section and HN9 and H10. The similarity of the GN (B) and CCN (A) R patterns for the 3–5 km altitude band is only due to the 0.996 R for N_{GN}-N_{1%} for these 5 flights (Figure 2 and Table 1, column 4). As described in the previous section this is the only reason for the positive R observed for N_{1%}-N_{d}. For all altitudes R for N_{1%}-N_{c} (left end of Figure 1a) is high whereas R for N_{GN}-N_{c} (left end of Figure 1b) is low, except 3–5 km. However, N_{GN} should not influence N_{c} any more than N_{1%} can positively influence N_{d}. Order of magnitude differences between N_{1%} and N_{d} and between N_{GN} and N_{c} are consistent with this (Table 1, columns 5–10). It is only the high N_{GN}-N_{1%} R for these 5 flights that forces equal R for N_{1%} and N_{GN} with anything. The pattern of high positive R for N_{1%} with N_{t} at large size thresholds when N_{CCN} are in the same proportions at all S was observed and predicted by HN9 and H10. The proportionality of N_{GN} with N_{1%} for this 3–5 km band is an example of this proportionality.

[9]Table 1, column 4 shows that N_{GN} and N_{1%} were uncorrelated for the other cloud altitude bands. That decoupling of N_{1%} from N_{GN} is reflected in the R differences between Figures 1a and 1b. Figure 1a shows negative R for all cumulative drop sizes larger than 40 μm for 5 of the 7 altitudes whereas Figure 1b shows positive R for all cumulative sizes above 120 μm at all altitudes. The positive R drizzle exceptions in Figure 1a are an artifact for the 3–5 km band (N_{GN} ∝ N_{1%}) and very small positive R for the 600–900 m band, which has least N_{d} (Table 1, columns 8–10).

[10]Figure 3 demonstrates the contrasting CCN and GN R patterns within each altitude band. For drizzle (right side of Figure 3) these R differences progressively increase for higher altitude bands except the 3–5 km band. This progressive increase with altitude of the differences between the drizzle R values of GN and CCN and the increase of R for N_{GN}–N_{d} with altitude, parallel the increase in N_{d} with altitude (Table 1, columns 8–10). The last column of Table 1 quantifies the R differences between N_{GN}-N_{d} and N_{1%}-N_{d} by altitude.

[11] The effects of CCN and GN on precipitation are opposite; precipitation is favored by higher N_{GN} while higher N_{CCN} impedes precipitation. This suggests that the most direct opposition of these influences when N_{GN} ∝ N_{1%} would convolute these influences. Although this might impair deconvolution of these influences by some types of analyses, here the monotonic N_{GN}-N_{1%} relationship offers the most direct comparison of these opposing influences. For the highest RICO altitude band the N_{GN} positive influence on N_{d} clearly dominates the negative influence of N_{1%} on N_{d}. The low N_{GN}-N_{1%} R (Table 1, column 4) of other altitude bands can be converted to this most direct comparison by considering only flights that have a monotonic N_{GN}-N_{1%} relationship. For the 2^{nd} highest altitude band (2.4–3 km) 4 of the 6 flights; (Figure 2 and Table 1, row 8) have a high N_{GN}-N_{1%} R (column 4). They display a higher positive N_{GN}-N_{d} R and significance level (Figure 4a) than N_{GN}-N_{d} for all 6 flights (Figure 3b). Figure 4a shows the same compulsory concurrence of the two sets of R as Figure 3a. Since N_{CCN} should only have a negative influence on N_{d}, this further demonstrates the dominance of the GN influence on drizzle at high cloud altitudes where there is more drizzle (Table 1, columns 8–10). N_{1%}-N_{d} is negative for all altitude bands where N_{GN} is not proportional to N_{1%} except the small positive R for the lowest band where there is least N_{d}.

[12] Two altitude bands had cloud data for the same 5 flights with clouds above 3 km (last two rows of Table 1). The R patterns for these considerations where N_{GN} ∝ N_{1%} are in Figures 4b and 4c, which demonstrates dominance of the negative influence on N_{d} by N_{1%}. The greater negative influence of N_{1%} on N_{d} than the positive influence of N_{GN} on N_{d} at lower altitudes is emphasized by the significantly greater negative R in Figure 4b than Figure 3d and Figure 4c compared to Figure 3g. This analysis demonstrates that the direct aerosol competition shown in Figure 3a and Figure 4 provides a more definitive assessment of the relative influences of the two aerosols on N_{d}.

4. Discussion and Conclusions

[13] The RICO flight plans were designed to capture clouds in early precipitation stages. The progressive increase of N_{d} with altitude (Table 1, columns 8–10) indicates success in that pursuit. The consistent positive R between N_{GN} and N_{d} progressively increases with altitude and size while the consistently negative R between N_{1%} and N_{d} mostly decreases in magnitude with altitude and size. This describes greater GN influence on larger drizzle drops at higher altitudes where there are more drizzle drops. Although the number of flights (data points) is smaller for the higher altitudes where R for N_{GN}-N_{d} is highest, the statistical significance levels are 95% for the two highest altitude bands and exceed 90% for the third and fourth highest altitudes. Although RICO fortuitously provided wide aerosol and cloud microphysics variations, it was not a controlled experiment. R for N_{CCN} or N_{GN} with N_{c}, N_{t} or N_{d} is reduced by other variables namely vertical velocity (W) and entrainment, which independently influence cloud microphysics.

[14] The N_{GN} used in this analysis (e.g., Table 1, column 6) are not exactly the particles upon which most of the drizzle drops condensed; i.e., different drizzle concentrations at various drop sizes and altitudes condensed upon different particle size or S_{c} ranges. But CR6 indicated that N_{GN} in all size ranges are probably in similar proportions among the flights because they all come from the same wind driven ocean source. Thus, the actual N_{GN} used in this analysis are probably reasonable surrogates for N_{GN} at various sizes or S_{c} that the various drizzle drops condensed upon; some may be smaller than GN.

[15] This analysis assumes that CCN and GN measurements well below cloud base are relevant up to 5 km. This is indicated by the consistently high R for N_{1%}-N_{c} (H9 and Figures 1, 3, and 4) and the very high R between N_{GN} at 100m [CR6] and 450m [Reiche and Lasher-Trapp, 2010] altitude. The wider range of altitudes within the higher altitude bands is a weakness, but there is no correlation between each average flight cloud altitude or average LWC with N_{d}.

[16] The N_{GN} correlation with surface wind is a complicating factor because of the horizontal wind coupling with W, which independently influences cloud microphysics. CR6 found low altitude cloud W correlated with horizontal wind velocity, R = 0.79 for 12 flights. This R is 0.90 for the 5 flights of the 3–5 km band. However W at the highest altitudes was not correlated with low altitude W or with low altitude horizontal wind velocity. This argues against the possibility that dynamic effects were the cause of the apparent GN influence on high altitude drizzle.

[17] This observational work will be expanded to include CCN spectra (i.e., intermediate of 1% and GN), comparisons between aerosol and drop number concentrations, and droplet growth model predictions (i.e., H10) that will provide more explanations. Although correlation is not cause, this analysis indicates viability of the GN warm rain hypothesis. Recent RICO articles by Gerber et al. [2008] and Lowenstein et al. [2010] also lend some support to the GN warm rain hypothesis. This analysis also has implications for the geoengineering proposal to counteract global warming by brightening clouds because injection of particles of too large a size range could have opposite effects such as those of deliberate hygroscopic cloud seeding [Bruintjes, 1999], which is supported by these results. This study indicates the importance CCN spectra not just total CCN concentrations.

Acknowledgments

[18] Support was from the U.S. National Science Foundation grant ATM-0342618. Cloud measurements were provided by the Research Aviation Facility of NCAR, which provided the C-130 airplane, the platform for all measurements.