CCN and cloud droplet concentrations at a remote ocean site

Authors


Abstract

[1] Extensive aircraft measurements over the mid Pacific displayed unexpectedly high overall average concentrations of condensation nuclei (CN) of 379 cm−3, cloud condensation nuclei (CCN) at 1% supersaturation (NCCN) of 301 cm−3, and total cloud droplets (Nc) of 139 cm−3. Factor of two variations in average NCCN and Nc among 11 flights produced a positive correlation coefficient (R) of 0.81. Updraft velocity (W) was also positively correlated with Nc but with lower R. Comparisons with a more traditionally maritime project (RICO), which had higher average W clearly demonstrated the role of W in determining Nc. Correlations of NCCN with concentrations of larger droplets were negative but the R switched back to positive for concentrations of largest cloud droplets with NCCN. The explanation offered for this double reversal of R for NCCN with droplet concentrations greater than various sizes suggests yet another influence of the aerosol on cloud microphysics.

1. Introduction

[2] The indirect aerosol effect (IAE) is the largest global climate uncertainty [Intergovernmental Panel on Climate Change, 2007]. IAE is due to anthropogenic cloud condensation nuclei (CCN), which are most abundant in polluted air masses, i.e., urban air [Squires, 1966; Frisbie and Hudson, 1993]. This usually gives rise to significant differences in CCN and total cloud droplet concentrations (Nc) between continental and maritime air masses [Squires, 1958]. Nonetheless, there has also been long range transport of continental/polluted air masses to maritime locals that also affected cloud microphysics [e.g., Garrett and Hobbs, 1995]. Here we report what appears to be an extreme example of persistent continental CCN and Nc at a remote maritime location. These observations have important implications for understanding and quantifying IAE. The results also present further evidence of the abundant influence of CCN on cloud microphysics.

2. Measurements

[3] The measurements were made on all 14 National Center for Atmospheric Research (NCAR) C-130 research flights between August 9 and September 7, 2007 during the Pacific Sulfate Experiment (PASE). Thirteen of the flights were more than eight hours in duration. Twelve of the flights were done during daylight and half of the other two flights were in morning darkness. Most of the time for all PASE flights was spent within the low boundary layer. All measurements were made within 500 km of and northeast of Kiritimati, Kiribati (Christmas Island of the Northern Line Islands), which is 230 km north of the equator and 2140 km directly south of Honolulu, Hawaii. The PASE location was chosen both for remoteness and dryness; Christmas Island is in an equatorial dry zone. Because PASE was a clear air chemistry project the season was even chosen for minimal cloudiness. Indeed clouds were small, few, and far between during PASE, but they could not be avoided on 11 flights (Table 1). Clouds were penetrated during both night flights (Aug. 21, RF6 and Sep. 5, RF13); the 4 hour afternoon flight of (Aug. 24, RF7) was cloud-free.

Table 1. Flight-Average and Standard Deviations (SD) of Concentrations (number per cm3) of Total Particles (NCN), CCN at 1% S (NCCN), and Total (diameter > 2.4 μm) FSSP Cloud Droplet Concentrations (Nc) for Parcels With More Than 0.1 gm−3 LWCa
NumberDateNCNNCCNNc
  • a

    Also shown is flight number and date.

RF1A09449 ± 12430 ± 43199 ± 51
RF2A11321 ± 25259 ± 29107 ± 32
RF3A14419 ± 11333 ± 63166 ± 55
RF4A16330 ± 40271 ± 45127 ± 42
RF5A19329 ± 26300 ± 32
RF6A21377 ± 13296 ± 34129 ± 38
RF7A24401 ± 40303 ± 25
RF8A26329 ± 26211 ± 3299 ± 33
RF9A28435 ± 7274 ± 25147 ± 48
RF10A30329 ± 21
RF11S02382 ± 12323 ± 27129 ± 36
RF12S03440 ± 37328 ± 32176 ± 37
RF13S05440 ± 31359 ± 38128 ± 33
RF14S07332 ± 8223 ± 39123 ± 20
Average 379 ± 22301 ± 36139 ± 39
SD 515830

[4] Table 1 shows that condensation nuclei (CN—total particle, with a TSI 3010 CNC) concentrations (NCN) and CCN [Hudson, 1989] concentrations (NCCN at 1% supersaturation [S]), and Nc (cumulative concentrations larger than 2.4 μm diameter with a PMS Forward Scattering Spectrometer Probe (FSSP; 2–46 μm diameter) were higher than expected for such an isolated location over the mid Pacific. These represent hours of data recorded every few seconds. Standard deviations (sd) in Table 1 that ranged from 2–12% of the NCN means (average 6%) and 8–19% of the NCCN means (average 12%) indicate uniform concentrations within the low boundary layer throughout each flight. During PASE the Intertropical Convergence Zone (ITCZ) was north of the research area and thus the predominant wind direction was east southeast. Back trajectories using the NOAA Air Resources Laboratory HYSPLIT transport and dispersion model and READY website (http://www.arl.noaa.gov/ready.html) from all 14 PASE flights extended to the vicinity of the west coast of South America at approximately 12 days.

3. Results

[5] Table 2 compares PASE with more typical maritime concentrations measured with the same instruments (calibrated on each and every flight) and airplane in the 2004–05 Rain in Cumulus over the Ocean (RICO) Eastern Caribbean project [Rauber et al., 2007; Hudson et al., 2009, hereinafter referred to as H9]. CN and CCN data in Table 2 are averages and sd of flight averages of extensive lower boundary layer measurements. The cloud microphysics measurements are also averages and sd of the averages of each research flight in each project. All RICO cloud data here are from the lowest altitude band (600–900m), which should be most similar to the shallow PASE clouds. The 45% higher average NCN and factor of 3 greater average NCCN in PASE were opposite of expectations for this more isolated region over the middle of the largest ocean. Figure 1b shows that the lowest PASE flight-average NCCN exceeded the highest RICO flight-average NCCN. The fact that average Nc was only 56% higher in PASE than RICO despite the 184% higher NCCN was probably largely due to the factor of two weaker average updrafts (W) (last column, Table 2) in PASE. Another reason that Nc in PASE was not as much higher as NCCN was in PASE compared to RICO may have been that the smaller PASE clouds were more susceptible to dilution and evaporation of entrainment. The difference in cloud size was most pronounced in the vertical where RICO clouds extended to more than 3 km (H9) while PASE clouds were never more than a few hundred meters thick. Nc was higher in PASE but cloud droplet mean diameter (MD) was higher in RICO because the ratio of average cumulative droplet concentrations of PASE to RICO steadily decreased with increasing threshold sizes. By 20 μm diameter this ratio reversed (RICO higher) and at 35 μm diameter RICO was a factor of three higher than PASE (data column 4, Table 2). At higher RICO altitudes concentrations of large droplets were even higher while Nc was constant with altitude. Lower concentrations of large droplets in PASE than RICO can be attributed to both greater competition of higher Nc and less cloud depth in PASE. The average 18% smaller MD in PASE also made PASE clouds more susceptible to evaporation reductions of concentrations at all sizes. All of these differences in macro and microstructure contributed to the absence of drizzle in PASE compared to significant drizzle in RICO (H9).

Figure 1.

Flight-averaged total cloud droplet concentrations (Nc) using cloud threshold 0.1 gm−3 LWC versus; (a) flight-averaged boundary layer CCN concentrations at 1% supersaturation (NCCN) for PASE, (b) same for PASE and RICO, (c) NCCN multiplied by square root of updraft velocity (W). Flight numbers (Table 1) are used as data points in Figure 1a. RICO data in Figure 1b is identical to Figure 1 of H9. For RICO only the 600–900m altitude band is considered. Linear regression equations and correlation coefficients (R) are shown. Two PASE data points are missing in Figure 1c because W was not available for RF9 and the W average was negative for RF4.

Table 2. Average and SD of Flight-Average Aerosol and Cloud Microphysicsa
 NCN (cm−3)NCCN (cm−3)Nc (cm−3)N > 35μm (cm−3)MD (μm)W (m/s)
  • a

    N > 35 μm is concentration of cloud droplets larger than 35 μm, MD is mean FSSP diameter, W is updraft velocity. Cloud data are with cloud threshold LWC > 0.1 gm−3. RICO cloud data are for only the lowest altitude band 600–900m (H9).

PASE379 ± 51301 ± 58139 ± 300.01212.9 ± 1.40.60 ± 0.40
RICO264 ± 73106 ± 3689 ± 360.03415.2 ± 1.91.13 ± 0.25

[6] Although low boundary layer aerosol concentrations were uniform within each PASE flight, there were significant concentration differences among the PASE flights. Table 1 shows that although flight-averaged NCN varied by only 37%, both NCCN and Nc varied by a factor of two among the PASE flights. Although this was only half of the factor of four variability of NCCN and Nc in RICO (H9), this was still enough variability to obtain the significant correlation shown in Figure 1a. This 0.80 correlation coefficient (R) has a significance level well above 99%, but this is not as high as RICO (Table 3 and H9) probably because of the smaller number of PASE cloud flights (11 versus 16) and only a quarter to a seventh as much cloud time (last column, Table 3). The factor of two smaller NCCN and Nc range in PASE than RICO and the greater entrainment in the smaller PASE clouds also contributed to the lower NCCN-Nc R of PASE.

Table 3. Comparisons of the Two Projects for Three Cloud Thresholdsa
LWC Thresh (gm−3)R NCCN-NcR NCCN-20μmR NCCN-35μmR W-NcR NCCNxW-NcR NCCNxW.5-NcNc Average (cm−3)Total time (s)
  • a

    R is correlation coefficient for flight averages. Symbols defined in Table 2. Last column is number of seconds of cloud measurements.

PASE
0.010.70−0.470.220.500.690.71812155
0.100.80−0.380.690.460.670.82139483
0.200.85−0.410.71−0.210.080.41159171
 
RICO
0.010.88−0.36−0.160.750.910.92568674
0.100.85−0.62−0.400.440.850.87892897
0.200.80−0.68−0.570.240.790.821011155

[7] Figure 1b shows a good NCCN-Nc R for combined project data, but the effect on Nc of the W and entrainment difference between projects can be noticed from the overlapping Nc and lack of overlapping NCCN between projects. The lower W of PASE indicates lower cloud S; so correlations with Nc might be better for NCCN at lower S. Greater competition among droplets may also have lowered the PASE cloud S. This suggests a higher order regression in Nc for Figure 1b; second order R = 0.82; third order R = 0.85.

[8] W also determines Nc, but R for W with Nc (column 5, Table 3) in all cases is lower than R for NCCN-Nc and shows greater variability with LWC. This R is significantly superior for RICO than PASE where it is even negative for 0.01 gm−3 threshold. When NCCN and W are considered together by correlating their products (NCCNxW) with Nc (column 6, Table 3), R for RICO is the same as for NCCN-Nc, but for PASE this R is intermediate of the R for W-Nc and NCCN-Nc. The significantly greater R for NCCN-Nc than W-Nc, which indicates the greater influence of NCCN than W on Nc, can also be simulated by correlating the products of NCCN and the square roots of W with Nc (column 7, Table 3). This NCCNxW0.5-Nc R is the highest R for RICO, but for PASE though this R is higher than the NCCNxW-Nc R it is lower than R for NCCN-Nc. Nonetheless, Figure 1c shows that this NCCNxW0.5-Nc regression for the combined data has a higher R of 0.85 compared to the NCCN-Nc R of 0.81 in Figure 1b. The Figure 1c accounting for W brings the data from the two projects closer together than in Figure 1b. But there is still an obvious significant difference in the separate linear regressions for each project in Figure 1c that could be caused by the greater entrainment or the lower cloud S in PASE.

[9] Figure 2a and column 3 of Table 3 show that R reverses for NCCN with cumulative concentrations of larger cloud droplets. This R reversal is probably due to competition among droplets for condensed water, which restricts droplet sizes to a greater extent when concentrations are higher. The lower magnitude of this negative R for PASE than RICO suggests less influence of NCCN on concentrations of larger droplets in PASE. Moreover, the high positive R for NCCN with the cumulative concentrations of the largest cloud droplets in PASE (column 4, Table 3) is very different from the negative R in RICO at these sizes. The large droplet-NCCN positive R values of PASE are probably related to the low droplet concentrations at the largest sizes; i.e., data column 4 of Table 2. Figure 2b shows the smaller average droplet sizes of PASE due to the higher NCCN and Nc. Thus, there was no growth of any high critical S [Sc] particles, which usually have high concentrations, out to the largest droplet size ranges for any of the PASE flights. Therefore, there was no significant competition for condensate in these large droplet size ranges. This is quite different from the situation at smaller droplet sizes where concentrations are sufficient to cause enough competition for condensate to result in greater limitations of droplet sizes in flights with higher NCCN. In PASE the large cloud droplets thus had grown on only the very low concentrations of large particles; i.e., only those with very low Sc (i.e., large or giant or even ultragiant nuclei). The PASE large cloud droplets are therefore not subjected to competition limitations and therefore their flight averaged concentrations retain a positive R with the aerosol concentrations upon which they condensed.

Figure 2.

(a) Correlation coefficients (R) for NCCN with FSSP concentrations of droplets larger than the various threshold diameters for PASE and RICO. LWC > 0.1 gm−3 is used to define cloud parcels. For RICO only the 600–900m altitude band is considered. (b) Average differential droplet distributions.

[10] Much of the time the relative concentrations of particles at different sizes are probably similar; i.e., when NCCN is higher the concentrations of large particles are also higher [e.g., Lasher-Trapp et al., 2008]. The positive R for PASE flight-averaged NCCN with cumulative droplet concentrations larger than 35 and 40 μm suggests this (Figure 2a). Furthermore, the R of the 35 μm flight-averaged droplet concentrations with flight-averaged boundary layer particle concentrations larger than 7 μm diameter measured with the FSSP is 0.73. This comparison is important because these two concentration ranges are nearly identical; the averages of these flight-averaged concentrations are within 50% of each other. Thus the largest cloud droplets may have condensed on roughly these same particles depending on their hygroscopicity. Concentrations of these particles were also positively correlated with NCCN though with a lower R.

[11] The preceding explanation also accounts for the lower magnitude negative R for PASE compared to RICO at intermediate sizes (i.e., 20 μm) and the reduced magnitude negative R for NCCN with large droplets in RICO. In these situations less advanced cloud droplet growth diminished the negative R tendency due to vapor competition, which then less effectively vied with the primary positive R tendency between aerosol and drop concentrations. This results in intermediate R values. At higher altitudes in RICO where large cloud droplet concentrations are one to two orders of magnitude higher, R is more negative out to larger sizes (H9) because higher concentrations at any size range are more sensitive to vapor competition limitations. At lower altitudes where droplets are smaller, the primary positive R tendency is realized down to smaller sizes and this counteracts and thus reduces the magnitude of the negative R caused by vapor competition of the numerous small droplets grown on the numerous higher Sc particles.

[12] The counteracting factors just described also explain the differences in R at the smallest droplet size ranges in Figure 2a. At 2 and 5 μm R is very high for both projects because for most of the flights most of the cloud droplets have grown larger than 5 μm and thus have grown on as many CCN as will be activated during those flights. In RICO R at 10 μm is just as high as at 5 μm because on most of the flights most of the droplets have grown larger than 10 μm (Figure 2b). R at 10 μm in PASE is lower because on most flights many droplets have not grown beyond 10 μm because the clouds are smaller and the concentrations of these droplets are limited by competition for condensate, which is what makes R negative. The negative R tendency is most fully realized at 20 μm, which is just enough larger than the mode of the droplet distribution (Figure 2b) to be most sensitive to the size limiting effect of vapor competition. The magnitude of R at 20 μm is smaller in PASE because the smaller droplet sizes cause the negative R tendency to less effectively vie with the positive R tendency that is fully manifested at either end of the cumulative droplet size range (Figure 2a). The larger droplet sizes of RICO produce a deeper negative R and maintain negative R out to larger sizes than PASE.

4. Comparisons

[13] CCN measurements by Twomey and Wojciechowski [1969] at many locations around the world showed maritime concentrations at 1% S below 200 cm−3. This was also reported by Hudson and Xie [1999] for extensive measurements in three maritime projects. Lists of many published cloud measurements compiled by Miles et al. [2000] reported maritime Nc below 200 cm−3. On the other hand Garrett and Hobbs [1995], Hudson and Li [1995], Hudson and Xie [1999], Heymsfield and McFarquhar [2001], and Hudson and Yum [2002] showed considerably higher NCCN and Nc over oceans. However, those observations were more limited in time and extent than PASE. The last two did report more than a month of polluted concentrations over the northern Indian Ocean, but this was within 1000 km of the very polluted subcontinent. Typical clean maritime NCCN and Nc were then consistently found south of the ITCZ in this INDOEX project.

[14] The very persistent high NCCN in the extremely remote PASE location so far from direct anthropogenic influences seems exceptional. Although the Nc values are not so shockingly high they were responsive to the NCCN variations. The lower W and greater entrainment in these small clouds probably restricted Nc to lower concentrations that did not reflect the NCCN. Nonetheless, the Nc were still higher than expected in such a remote area for such small clouds and the NCCN provided the potential for even higher Nc in conditions of greater W that could produce higher S. More extensive clouds would also be less susceptible to entrainment, which would also raise Nc beyond the PASE observations.

[15] Bodhaine and DeLuisi [1985] reported eight years (1976–1983) of NCN at American Samoa; 2350 km south southwest of Kiritimati and also within the southeast trade winds. With little annual variability the overall mean NCN was 274 cm−3, which is 72% of the PASE mean NCN and within 4% of the RICO NCN mean. Samoa monthly averages for August and September were 262 and 284 cm−3, but the 1981 averages for these months were 365 and 362 cm−3, which is within 5% of PASE. The overall NCCN to NCN ratio was 79% for PASE compared to 40% for RICO. This and the lower RICO concentrations may be due to the greater frequency and size of clouds in RICO. Cloud scavenging is a principal mechanism for reducing particle concentrations, especially CCN [Twomey, 1977] and in terms of cloudiness and precipitation, Samoa is more similar to RICO than to Kiritimati. Therefore, although NCN in PASE was not so much higher than at Samoa it seems likely that PASE NCCN was higher than at Samoa.

5. Conclusions

[16] The arbitrary but often quoted dichotomy between maritime and continental air masses based on particle (especially CCN) or cloud droplet concentrations of less than approximately 200 cm−3 now seems an oversimplification. The persistence of high particle and cloud droplet concentrations in a very remote part of the world was surprising. However, the concentrations noted in the preceding paragraph also suggest that at least CN concentrations in maritime air are more frequently higher than has often been assumed. Perhaps there is more natural production of particles or more long range transport of continental/pollution particles. The trajectories noted at the end of Section 2, although very uncertain, suggest that the possible transport of continental aerosol is over very long distances and time periods. Another possibility for previous underestimates of maritime concentrations is under appreciation for cloud scavenging. Since many maritime measurements, especially CCN, have been associated with cloud research, this may have biased data toward lower concentrations that had resulted from cloud scavenging. Since PASE was deliberately planned to avoid clouds and there were few and only small clouds during PASE, this may have led to higher CCN and cloud droplet concentrations than have typically been found in maritime air masses.

[17] Nonetheless, the few clouds that were found during PASE did respond to the different CCN concentrations (NCCN). The difference in updraft velocities (W) between PASE and RICO displayed the impact of W on droplet concentrations. However, NCCN variations had more of an effect on droplet concentrations than W variations within each project and when both projects were considered together. Both the low W and small sizes of the PASE clouds restricted the PASE droplet concentrations. The NCCN observed in PASE will have more impact on cloud microphysics in more vigorous cloud systems.

[18] Like RICO, PASE showed the strong positive correlations (R) between NCCN and total cloud droplet concentrations (Nc) for all cumulative or interval liquid water content (LWC) cloud thresholds. There was also a negative R with concentrations of larger/intermediate sized cloud droplets that was lower in magnitude than that in RICO. An explanation was advanced for this difference, which also elucidates the positive R for NCCN with the concentrations of largest cloud droplets in PASE and the lower R for NCCN with concentrations at some of the intermediate sizes in RICO. This explanation is one of conflict between a negative R tendency of vapor competition among high droplet concentrations within the mode of the droplet distributions with the positive R tendency outside of the mode; either beyond the mode (only largest droplets) or including the entire cloud droplet mode (total or small cumulative concentrations). This explanation suggests that there may be greater aerosol influence on cloud microphysics than can be indicated by the correlation coefficients. Thus, lower magnitude R for NCCN with cloud droplet concentrations does not always indicate less aerosol influence and thus greater influence of dynamics on cloud microphysics.

[19] This research is continuing and will next focus on the effects of the entire CCN spectrum. This will include predictions of Nc from CCN spectra and W compared with estimates of adiabatic (unentrained) Nc [Hudson and Yum, 2002]. This will better define cloud S and thus the CCN that actually make the cloud droplets. Vertical differences in NCCN will also be examined.

Acknowledgments

[20] The PASE work was supported by National Science Foundation (NSF) ATM-0627227. The RICO work was supported by NSF ATM-0342618. All measurements were made on board the NCAR C-130 airplane, which is also NSF supported.

Ancillary