Relationships between CCN and cloud microphysics variations in clean maritime air



[1] Flight-averaged CCN concentrations during the Rain in Cumulus over the Ocean (RICO) field project varied by a factor of four. This was sufficient to provide correlations with flight-averaged total cloud droplet concentrations (Nc) and mean diameters (MD) that showed greater influence of CCN than updraft velocity (w) on Nc and especially MD. The product of CCN and w produced even stronger correlations with Nc and MD. An earlier RICO study had shown an inverse relationship of Nc and giant nuclei (GN) with large cloud droplet concentrations (NLD). These two RICO studies now show that the aerosol with the most influence on NLD (MD) is CCN. The results of the present study are similar to earlier studies in more polluted clouds that also showed more influence of CCN than GN on MD/NLD/drizzle. The present study indicates that variations in CCN concentrations even within maritime air masses modulate precipitation.

1. Introduction

[2] RICO (Rain in Cumulus over the Ocean) (R. Rauber et al., Rain in cumulus over the ocean—The RICO campaign, submitted to Bulletin of the American Meteorological Association, 2006) investigated warm rain initiation in December (D) 2004 and January (J) 2005. It is well known that the concentration and spectra of the preexisting aerosol, namely cloud condensation nuclei (CCN), influences cloud droplet concentrations (Nc). Indeed along with the updraft velocity (w) at cloud base, CCN spectra determine initial Nc. However, Nc can subsequently be changed by mixing among cloud parcels, entrainment of out-of-cloud air (dilution and evaporation), and droplet coalescence. These factors are related to the first indirect aerosol effect (IAE) (cloud radiation) because many CCN are anthropogenic. Warm precipitation may also be influenced by CCN concentrations because MD, could or should be inversely related to Nc. This would relate to second IAE (precipitation inhibition). However, MD (NLD) may also be separately influenced by the concentration of giant (GN) or ultragiant nuclei (UGN).

[3] The differentiation of the effects on precipitation of the two aerosol classes, CCN and GN/UGN was an important goal of RICO and previous studies. During the Small Cumulus Microphysics Study (SCMS) on the East Coast of Florida in July–August, 1995 there were enough differences in wind direction to provide sufficient CCN and Nc differences to test that aerosol hypothesis. Hudson and Yum [2001] (hereinafter referred to as HY1) showed that variations in CCN concentrations not only determined Nc but also were inversely related to MD and drizzle drop concentrations. HY1 also presented evidence that there were insufficient variations in GN concentrations to produce the observed variations in drizzle. All of the SCMS CCN and Nc concentrations were characteristic of continental or polluted air masses; i.e., concentrations were definitely polluted with offshore winds and modified maritime with onshore winds. Other SCMS investigations of GN effects on precipitation were rather inconclusive [e.g., Laird et al., 2000; Blyth et al., 2003]. Therefore, RICO was planned for the island of Antigua at the northeast corner of the Caribbean Antilles, where predominant east–northeast winds have a long unpolluted Atlantic fetch. This suggested minimal CCN variability that might allow more definitive tests of the GN-MD/NLD/drizzle connection.

[4] During RICO there were enough differences in horizontal wind velocities (v) to produce significant differences in GN concentrations [Colon-Robles et al., 2006] (hereinafter referred to as C-R6). They found that the GN concentrations were inversely associated with NLD. Since NLD was also inversely related to Nc this might seem to support the CCN influence on precipitation. However, surrogates for CCN—Condensation Nuclei (CN) (0.01–3 μm diameter) and the Passive Cavity Aerosol Spectrometer Probe (PCASP) (0.14–2.75 μm diameter) particles were not correlated with v. Since C-R6 did find a good correlation between v and vertical wind (w) and between w and Nc, and higher NLD when w and Nc were low, they concluded that lower w, which should produce lower Nc, enhanced precipitation. Thus they determined that w was an important determinant of both Nc and precipitation in RICO. We now present RICO CCN measurements that demonstrate that CCN variability has a greater influence on Nc, MD and precipitation.

2. Measurements

[5] CCN were measured from the National Center for Atmospheric Research (NCAR) C-130 airplane by two Desert Research Institute (DRI) CCN spectrometers [Hudson, 1989]. We present data from the same 12 flights (RF3, 5–10, 12–15 and 18; Table 1) for which cloud data was shown by C-R6 and from Research Flights (RF1 and 2) of D7 and 8. Although the DRI CCN spectrometers provide complete spectra (2-0.02% supersaturation [S]) we confine this presentation to 1% S cumulative concentrations that were averaged over large portions of two different sets of two half-hour circles flown at 100 and 300 m altitude shortly after takeoff and just before return to base. Clouds were penetrated in the same area during the intervening periods of each flight.

Table 1. Flight Number, Date, Cloud Penetration Duration (s), Flight-Averaged Cloud Droplet Concentrations (Nc) (cm−3), CCN Concentrations at 1% S (cm−3), Updraft Velocities (w) (m s−1), and Mean Cloud Droplet Diameters (MD) (μm)a
FlightDateDurNc DRINcC-R6Nc DRI RankNcC-R6 RankCCN 100 mCCN 300 mCCN 100 m RankwW RankMDMD Rank
  • a

    The two Nc columns are those determined from this study and from C-R6. Adjacent Nc, w and MD columns show flight rankings of these quantities for the 12 flights. MD rankings are reversed; smallest 1, largest 12.

RF1D7185201   200198 1.77 13.8 
RF2D81123   133  1.64 15.4 
ave 39111   112112 2.03 17.1 
sd 4546   3535 0.28 2.1 

[6] CCN concentrations showed little variability over each of these low altitude circles except for occasional spikes that seemed to be remnants of ship exhausts (Figure 1). These were usually removed from the final data, as were occasional anomalous high concentrations during drizzle penetrations. After this editing the standard deviations (sd) of the concentrations over the 100-m circles ranged from 9–32% of the means with an average of 17% and sd of these sd of 5%. For the 300-m circles the sd range was 8–25% of the means with an average sd of 15% of the mean and sd of 5%. Some of this measured variability was caused by inherent counting limitations because even with constant concentrations, measurements should have an sd equal to the square root of the absolute number of counts within each record. The sample flow rate was typically 0.33 cm3 s−1 of which all CCN produced individual pulses. Since the integration time in Figure 1a was 4 s, 1.33 cm3 were sampled with each record. With a mean concentration of 41 cm−3 this meant that approximately 55 pulses were counted per record. Since the square root of 55 is 7.4 this implies an inherent sd of 13% of the mean. Since the measured sd was 34% of the mean there was real variability of the concentration that is obvious in Figure 1a. In Figure 1b the 3 s integration time processed 1 cm3 per record so on average 72 pulses were counted in each record. This implies an inherent sd of 8.5, which is similar to the measured sd of 10 cm−3. This means that apart from the spike, the actual CCN variability for this circle could have been less than the measured variability. A similar analysis shows that the variability may also have been overestimated in Figure 1c.

Figure 1.

CCN at 1% S with local time in 100-m circles. Data in red are peaks that were removed. (a) Dec. 17 morning, 240 records of 4 s duration; mean 41 cm−3; sd 14 cm−3. (b) Jan. 11 afternoon, 244 records of 3 s; mean 72 cm−3, sd 10 cm−3. (c) Jan. 23 morning, 268 records of 4.5 s; mean 132 cm−3, sd 17 cm−3.

[7] The differences between each 300 and 100-m circle done consecutively in the same location ranged from 8% lower to 9% higher with an average of 1% higher and sd of 5%. The afternoon concentrations averaged 19% higher than the morning concentrations with an sd of 44% and a range of 198 to 66% of the morning concentrations. For comparisons with the clouds, the averages of the two sets of circles were averaged for each flight and displayed in Table 1, which shows that flight-averaged CCN concentrations and Nc differed by a factor of 4 among the 14 flights.

[8] We used nearly the same criteria for obtaining flight-averaged Nc from the NCAR PMS Forward Scattering Spectrometer Probe (FSSP) as were used by C-R6; i.e., 600–900 m altitude, liquid water content (LWC) > 0.25 g m−3, w > 0.5 m s−1. We did not discriminate according to the PMS 260X probe (drizzle) and we used 1 rather than 10 Hz data. Nevertheless, except for RF9 and 15 with 8 and 7% Nc differences, we obtained average Nc that were within 4 percent of those displayed in Figure 3a of C-R6. In terms of flight-ranked Nc we both determined the same top three (RF14, 12, and 13) and bottom three (RF7, 8, and 12) in the same order and the same rank 7 (RF8). Ranks 8 and 9 are reversed (RF9 and 10) and ranks 4, 5, and 6 (RF6, 15, and 18) are scrambled.

3. Results

[9] The 0.70 correlation coefficient (R) for flight-averaged Nc versus flight-averaged CCN in Figure 2a indicates that CCN concentration variations had slightly more influence on Nc than flight-averaged w variations; R = 0.66 for Nc–w [C-R6]. The R for CCN on the 300-m circles was 0.68. Addition of two flights (RF1 and 2: D7 and 8) advances R to 0.80 for the 100-m circles (Figure 2a). RF7 (D17) wasn't the only flight with very low CCN concentrations; RF17 (J19) had concentrations of only 62 cm−3, but the clouds were too low (<600 m) and had insufficient LWC.

Figure 2.

(a) Flight-averaged cloud droplet concentrations (Nc) against flight-averaged CCN concentrations at 1% S at 100-m altitude for the 12 flights used by C-R6 plus RF1 and 2. Nc for altitudes 600–900 m, LWC > 0.25 g m−3, and w > 0.5 m s−1. (b) As Figure 2a but Nc versus the product of CCN concentrations and normalized flight-averaged updraft velocity (w). Linear regression equations and correlation coefficients (R) are shown for both the 12 flights (pts.) and the 14 flights. Regression lines are shown for 14 flights.

[10] The combined effects of CCN and w are very simply shown in Figure 2b. This brings R up to 0.82 for the 12 flights and 0.86 for the 14 flights. Since this exercise is only meant to estimate relative influences, these products were normalized by dividing by the average RICO w of 2.03 m s−1. This kept the same abscissa range. The coincidence of the abscissa and ordinate ranges in Figure 2 may be coincidental because the actual maximum cloud S values may have been higher than 1%, and Nc subsequently reduced by dilution, evaporation, or coalescence. The combined effects of w and CCN are also revealed by the flight rankings of CCN, w and Nc shown in Table 1. The R for (CCN*w)-Nc is 0.74 and 0.76 for C-R6 or DRI Nc rankings respectively. By comparison the w-Nc rankings have R of 0.55 and 0.60 respectively and the CCN-Nc rankings have R of 0.61 and 0.59 respectively. More sophisticated ways of combining these effects to predict Nc for comparison with measured Nc (dynamic CCN closure) involve the entire CCN spectra and guesses of the condensation coefficient [Yum et al., 1998].

[11] C-R6 noted a trend of increasing NLD with decreasing v and w. In Figure 3a, where the flight numbers are used as the data points, CCN are only weakly correlated with v, R = 0.37. C-R6 noted that RF6 and 18 (D16 and J23) were exceptions to the inverse relationship of NLD to v and w because they had low NLD in spite of 2nd and 3rd lowest v, moderate (6th ranked) and very weak (11th ranked) w and moderate Nc (5th or 4th ranked and 6th or 5th ranked), respectively. The explanation for these exceptions is that these flights had the 2nd and 3rd highest CCN concentrations (Table 1 and Figure 3a where they lie out to the upper left). When these two flights are excluded from the regression, R is 0.82 (Figure 3b). Therefore, since CCN are so well correlated with v for the ten flights for which NLD is inversely related to v and w, NLD is inversely related to CCN concentrations. The two exceptions are further testimony of the inverse effect of CCN on NLD. The 5th or 4th and 6th or 5th Nc rankings closely match the 4th and 7th ranks of the averages of the CCN and w of these two flights.

Figure 3.

All measurements at 100 m altitude: (a) CCN at 1% S versus horizontal wind speed, data points are flight numbers. (b) As Figure 3a but excluding RF6 and RF18 (D16 and J23). (c) PCASP concentrations versus CCN concentrations.

[12] Five of the ten flights that followed the NLD-v-w inverse trend are displayed in Figure 4 of C-R6 and Table 2 here in NLD order. The order of v and Nc for these flights is, of course, completely reversed, while CCN and w are mostly reversed. The flight with the highest NLD and largest MD (D17) had the lowest Nc and CCN but only the 3rd lowest w. The flight with the lowest NLD and smallest MD (J14) had the highest Nc and CCN but only the 2nd highest w. The last column, which combines CCN and w, is in complete reverse order.

Figure 4.

Flight-averaged plots. (a) cloud droplet mean diameter (MD) against cloud droplet concentrations (Nc). (b) MD versus CCN. (c) MD versus vertical velocity (w). (d) MD versus normalized product of CCN and w. All 14 data points are shown in Figures 4a, 4b, and 4d. Regressions are nearly identical for 12 data points. In Figure 4c, 12 data points are shown because R for 14 points is much lower.

Table 2. Flight Ranking of Large Cloud Droplet Concentrations (NLD) Denoted by C-R6 for Five of the Ten Flights That Followed the NLD-v Inverse Trend, Flight Number, Date, Horizontal Wind (v) (m s−1), and 12-Flight Rankings of Cloud Droplet Mean Diameter (MD) (Reverse Order), Cloud Droplet Concentration (Nc), Updraft Velocity (w) and Average of CCN and w
NLD RankFlightDatevMD RankNc DRI RankNcC-R6 RankCCN RankW RankAve of CCN and w Rank

[13] The fact that C-R6 found a non-correlation and even an inverse correlation of PCASP and CN with v and thus also apparently with Nc (Figures 1b and 1c of C-R6) suggested that CCN are less influential than w. However, Figure 3c here indicates that PCASP was not a good surrogate for CCN, R = 0.18. Moreover, PCASP concentrations also did not correlate with Nc; R = 0.17 or 0.14 for C-R6 or DRI Nc. This indicates that aerosol size measurements in RICO were not a useful surrogate for CCN as was recently suggested by Dusek et al. [2006].

[14] Figure 4a shows a very close relationship between Nc and MD (R = 0.94, second order R = 0.97). The linear coefficients of determination (R2) of Figures 4b and 4c show that CCN variations account for 65% of the MD variability while w accounts for only 35% of the MD variability. Figure 4d shows only slightly higher R for the combined effects of CCN and w on MD than CCN-Nc (Figure 4b). The visual and regression equation similarities of Figures 4b and 4a compared to Figures 4c and 4a demonstrate the greater influence of CCN than w on MD during RICO.

4. Discussion

[15] The major uncertainty that the low altitude CCN measurements represent the input to the clouds is the large differences between the morning and afternoon circle CCN concentrations (twice as much to two-thirds). When morning or afternoon CCN averages are considered alone the CCN-Nc R drops to 0.52 and 0.61 for 12/11 flights and 0.71 and 0.74 for 14/13 flights. The number of flights is lower for afternoon comparisons because J5 had no afternoon circle CCN measurements. All measurements were from the newer DRI CCN spectrometer except D13 (RF5) when the older DRI CCN instrument had to be used. But it operated at a lower S range that required extrapolation of its calibration curve to obtain 1% S measurements. This is less accurate than the calibration curve interpolations used for the newer instrument on all other flights. However, removal of J5 and/or D13 had nearly imperceptible effects on R.

[16] The small amount of cloud data (1s and 2s) for RF2 and 7 seems to reduce the value of these two data points in Figures 2 and 4. However, there is very little difference in the flight rankings and R values when the criteria for choosing cloud data are relaxed to expand the amount of cloud data. Exclusion of RF2 and/or RF7 from consideration has very little effect on any R in Figures 2 and 4 except when only 11 points are considered in Figure 2 (R = 0.61 A and 0.77 B).

[17] Predictions of Nc based on CCN and w can at best only estimate adiabatic Nc. Although the Nc criteria noted in the last paragraph of Section 2 were chosen for adiabaticity, they could not necessarily insure unaltered Nc, but the tight Nc-MD correlation does suggest adiabaticity (Figure 4a). It is possible, nevertheless, that the reductions from adiabatic Nc were usually in similar proportions and this produced ambient Nc that are proportional to adiabatic Nc. This could still result in good CCN-Nc correlations, especially if w is properly accounted for. Variations in w, however, should produce variations in cloud S, which would then require for Nc comparison, CCN concentrations at different S instead of all CCN concentrations at the same S (1%) that are considered here. However, actual cloud maximum S that determines Nc are due to the interaction of the entire CCN spectrum with w. Dynamic CCN closure should involve precise comparisons of the CCN spectra going directly into a given cloud parcel with measurements of that same cloud parcel, but comparisons of flight averages that must involve CCN at only one S (i.e., 1% here) may be more valuable from a climate perspective.

5. Conclusions

[18] We have shown that the factor of four variability in flight-averaged CCN concentrations observed during RICO influenced both the total cloud droplet concentrations (Nc) and the concentrations of large cloud droplets (NLD); e.g., cloud droplet mean diameters (MD). In combination with the C-R6 results this analysis shows that the aerosol with the most influence on Nc, MD and NLD (precipitation) is CCN. As noted by C-R6 these results tend to further confirm modeling studies of Ochs and Semonin [1979], Johnson [1982], and Feingold et al. [1999] that GN have little influence on precipitation when CCN concentrations are low.

[19] Yum and Hudson [2002] and HY1 found that CCN variations influenced Nc and precipitation in air masses with higher CCN and Nc concentrations; i.e., continental/polluted air masses. Those two studies also required much greater differences in CCN and Nc concentrations in order to display significant correlations. There were only good CCN-Nc or CCN-MD or drizzle correlations when different air masses were considered. There were not good correlations within only maritime or within only continental air masses. In the present study significant correlations are found within air masses that are all within the maritime/clean concentration range. The present results demonstrate that CCN variability modulates precipitation in clean maritime air masses.

[20] This analysis shows that both variations in CCN concentrations and updraft velocity (w) affect Nc and precipitation, with CCN exerting the greater influence. Taken in combination the results of all of the publications noted in this section tend to uphold the viability of both the first and second indirect aerosol effects (IAE) (still the largest climate uncertainty [Intergovernmental Panel on Climate Change, 2001]) because many CCN are anthropogenic. Results here indicate that maritime clouds are not only especially susceptible to first but also to second IAE (precipitation inhibition), i.e., changes in low CCN concentrations affect precipitation. More thorough analyses of RICO CCN, dynamics, and cloud microphysics will follow in other journals.


[21] Support was from the US National Science Foundation Grant ATM-0342618. Measurements other than CCN were provided by the Research Aviation Facility of NCAR, which provided the measurement platform, the C-130 airplane.