Geophysical Research Letters

Stratus cloud supersaturations

Authors


Abstract

[1] A wide range of California coastal stratus cloud droplet concentrations (Nc) were compared with CCN spectra to provide numerous estimates of cloud supersaturations (S). These revealed the predicted decrease of S with higher CCN concentrations (NCCN). A significant result was frequent stratus cloud S significantly above 0.3%, especially for more maritime NCCN and Nc (i.e., <300 cm−3). This means that much smaller and more numerous particles than previously considered interact with the most important clouds for the indirect aerosol effect.

1. Introduction

[2] The largest climate uncertainty, the indirect aerosol effect (IAE) [Alley et al., 2007], is based on large variations of observed cloud microphysics. IAE is mostly due to purported alterations of stratus clouds, which have a pivotal climate role because they cover large expanses of ocean where they provide an immense albedo contrast with the bare ocean. Low altitude maritime clouds provide high radiative temperatures that result in global cooling. Increases in stratus cloud droplet concentrations (Nc) due to higher concentrations of cloud condensation nuclei (CCN) enhance cooling by stratus clouds. The accompanying decrease of droplet sizes inhibits precipitation, which then often enhances the spatial and temporal extent of these clouds. Both of these effects of higher CCN concentrations (NCCN) due to air pollution constitute IAE. The usual low NCCN and low Nc in maritime air masses makes these clouds more susceptible to anthropogenic influences [Platnick and Twomey, 1994]. However, it has generally been thought that low vertical velocities (W) observed in stratus clouds result in lower supersaturations (S), which means that only low critical S (Sc) CCN produce stratus cloud droplets. This smaller CCN subset generally consists of larger particles. Based on many observations [e.g., Hudson, 1983; Hoppel et al., 1986; Leaitch et al., 1996; Roberts et al., 2006] Hegg et al. [2009] reaffirmed the long held conventional wisdom that stratus S are < 0.3% (100.3% RH).

[3] Initial Nc are determined by CCN spectra, temperature and W at cloud base. It has long been known that the greater competition for condensate with higher Nc caused by higher CCN concentrations (NCCN) should restrict cloud S [i.e., Twomey, 1959]. However, this phenomenon has seldom been observed in stratus [Hudson and Yum, 2002] or any other types of clouds. Hudson et al. [2010] recently reported lower S with higher NCCN in higher altitude continental clouds.

2. Measurements

[4] The field experiment Physics of Stratocumulus Tops (POST) in July–August, 2008 off the central California coast provided a wide enough range of measured NCCN and Nc to observe systematic variations in stratus S. Unlike recent presentations by the authors (Hudson et al. [2009], RICO (Rain in Cumulus over the Ocean); Hudson and Noble [2009], PASE (Pacific Aerosol Sulfate Experiment)), POST provided both intraflight and interflight contrasts of NCCN and Nc so that 15 research flights provided an order of magnitude more NCCN-Nc comparisons than RICO or PASE.

[5] All measurements were made on board the CIRPAS Twin Otter airplane based at Marina, California. Condensation nuclei (CN) were measured with a TSI 3010, CCN were measured by a Desert Research Institute CCN spectrometer [Hudson, 1989] that was calibrated more than once during each flight, cloud droplets were measured with a Cloud Aerosol Spectrometer (CAS) probe (diameter 0.58–51 μm), W was measured with a GPS corrected C-MIGITS III using the technique of Lenschow and Spyers-Duran [1989].

[6] Table 1 characterizes the 9 hours of cloud and CCN measurements (4.5 hours each). The first five columns are 69 vertical slant penetrations completely through the cloud deck from below the bases to above the tops of the stratus. The last two columns characterize 28 horizontal cloud passes through solid continuous cloud and their corresponding below cloud CCN measurements.

Table 1. Altitudes of Cloud Base, Top and Thickness, and Durations of Clouds and Corresponding Below Cloud CCN Measurementsa
 Base (m)Top (m)Thick (m)Vertical Cloud (s)Vertical CCN (s)Horizontal cloud(s)Horizontal CCN (s)
  • a

    Mean, standard deviation, minimum, maximum and median altitudes in meters, and durations in seconds. First five columns are for the 69 vertical cloud penetrations; last two columns are for the 28 horizontal cloud penetrations. LWC > 0.1 gm−3 for all clouds. Last row shows total altitude cloud thicknesses and total of the durations.

Mean191513321143165227184
SD1221387777127136143
Min272811436148015
Max510976606361567475567
Median164483312118150179150
Total  2216698901138063665143

[7] The presented Nc are average concentrations of droplets larger than 0.58 μm diameter for those one second periods when CAS liquid water content (LWC) exceeded 0.1 gm−3. The presented NCCN are simultaneous averages over the S range of 0.04–1.5% over horizontal flight legs below the clouds. For the vertical cloud penetrations CCN were measured immediately before the ascents or after the descents. CCN measurements associated with horizontal cloud penetrations were often more separated in time and distance from the cloud measurements.

3. Results

[8] Figure 1 displays mean Nc of each vertical penetration as a function of the corresponding below-cloud NCCN, here N1% (at 1% S) is used because it includes most CCN and N1% has often been used to characterize air masses. The wide range of concentrations, the low slopes of the regressions and the nonlinearity of the response of Nc to NCCN is evident. Although the influence of CCN on cloud microphysics is generally thought to be due to CCN from below the clouds there are often very high NCCN immediately above California coastal stratus [e.g., Hudson and Frisbie, 1991]. In an attempt to ascertain the possible direct influence of the often very different above cloud NCCN, the 69 vertical cloud passes are divided according to vertical aerosol profiles. CN concentrations (NCN) are used for this purpose because the fast response of CN instruments provides better discrimination from within cloud measurements, which are often subject to splashing artifacts. This is feasible because NCN were generally proportional to NCCN. Figure 1 indicates little difference in the Nc-N1% relationship between the 40 cases with higher NCN immediately above the stratus than below the stratus (designated HA) and the 29 cases with just above cloud NCN lower or similar to below cloud NCN (designated LSA). The high and similar correlation coefficients (R) for Nc-N1% and the fact that this R is higher for the 40 HA cases than the 29 LSA cases suggests that there was minimal direct influence of above cloud CCN on Nc. This is also indicated by the results of a division of each cloud penetration into altitude quartiles. This revealed the highest Nc-N1% R for those Nc measured in the highest cloud altitude quartile. The Nc-N1% R values in Table 2 are similar to those for Caribbean small cumuli of RICO [Hudson et al., 2009] and small mid-Pacific trade wind clouds of PASE [Hudson and Noble, 2009].

Figure 1.

Average total cloud droplet concentrations (Nc) during 69 vertical penetrations of stratus versus corresponding below cloud CCN concentrations. Blue–29 clouds with lower or similar CN concentrations (NCN) above than below cloud, LSA; red– 40 clouds with higher NCN above than below cloud, HA. Linear regressions are shown for each set and 2nd order regression for all 69; R-correlation coefficient, R1 linear, R2 2nd order.

Table 2. Correlation Coefficients, R, Between Various Parameters for the Vertical and Horizontal Cloud Penetrationsa
 VerticalHorizontalVertical HAVertical SAVertical Adiabatic
  • a

    Vertical penetrations are divided according to higher (HA), or lower and similar (LSA) NCN above each cloud compared to below each cloud. Vertical clouds with adiabatic LWC altitude profiles are also separated. First row is number of clouds in each group. All cloud data are for 0.1 gm−3 threshold. Nc-mean total cloud droplet concentrations, N1%-below cloud CCN concentrations at 1% S, W-within-cloud vertical velocity, Seff-effective cloud S as defined in Figure 2 and text.

Number6928402917
Nc-N1%0.850.860.890.840.79
Nc-W0.450.510.390.51−0.23
N1%-W0.390.510.300.53−0.28
Seff-N1%−0.74−0.83−0.86−0.35−0.84
Seff-Nc−0.35−0.51−0.620.13−0.41
Seff-W−0.12−0.44−0.06−0.18−0.57

[9] Figure 1 shows similarity between Nc and N1% for N1% < 175 cm−3, but Nc a factor of 2 lower than N1% for N1% > 500 cm−3. This suggests the lower cloud S for higher NCCN predicted by Twomey [1959]. Figure 2 quantifies this phenomenon by comparing each average CCN spectra (1.5 − 0.04% S) with each corresponding mean Nc of the 69 vertical cloud penetrations. The S for which NCCN most closely matches mean Nc is Seff [i.e., Hudson, 1983]. Seff was set to 1.6% for the 5 clouds with mean Nc > N1.5%, and to 0.03% for the 3 clouds with mean Nc < N0.04%. Although the data are quite scattered, the R values are substantially negative indicating suppression of cloud S at higher NCCN. As in Figure 1 the R for the 40 HA clouds is of greater magnitude than R for the 29 LSA clouds. Although again this suggests minimal direct influence of the above cloud aerosol on Nc, this R difference is probably a result of the more limited N1% upper range for the 29 LSA clouds compared to the 40 HA clouds.

Figure 2.

As Figure 1 but plotting the effective supersaturation (Seff) against below cloud CCN concentration (NCCN). Seff is the S for which NCCN matches Nc. Only linear regressions are displayed.

[10] Figure 3 displays Nc-N1% for the 28 horizontal cloud parcels. In spite of horizontal concentration variations during most of the flights, measurements over longer cloud paths (Table 1) and greater distances between the CCN and the cloud measurements, R is similar for the 28 horizontal clouds and the 69 vertical clouds (Table 2). Figure 3 also displays the Nc-N1% relationship for the 17 vertical clouds with apparent adiabatic LWC altitude profiles. This is one attempt to minimize the effects of entrainment and droplet coalescence that reduce Nc and thus obscure relationships between Nc and NCCN that generally establish Nc at cloud base. Since such reductions of Nc could be independent of NCCN they could perturb the Nc-NCCN relationship and thus diminish the relevance of Seff that is based on Nc. The most relevant Seff is that which reflects the influence of NCCN on Nc at cloud base. Entrainment and coalescence usually also reduce or move LWC, which thus disrupts the linearity of adiabatic LWC altitude profiles; over small stratus depths, adiabatic LWC is approximately linear. Therefore, those vertical penetrations that displayed linear LWC altitude profiles ought to show higher R for Nc-NCCN than R for clouds with nonlinear LWC altitude profiles. The 17 linear LWC profiles also agreed with calculations of adiabatic LWC based on cloud base temperature and pressure. However, R of Nc-N1% for the 17 adiabatic clouds is lower (Figure 3 and Table 2). This may be due to this small number of samples or to the more limited upper range of N1% for the 17 adiabatic clouds. The fact that R is not greater for the adiabatic clouds may also be due to the fact that the reductions of Nc in the sub-adiabatic clouds do not significantly disrupt the Nc-N1% relationships; i.e., the reductions of Nc may be in proportion to N1%.

Figure 3.

As Figure 1 but for 28 horizontal cloud passes and for 17 of the vertical penetrations shown in Figure 1 that had linear liquid water content (LWC) profiles that approximated predicted adiabatic LWC profiles.

[11] Figure 4 shows stronger Seff-N1% relationships for the horizontal cloud legs and the adiabatic vertical clouds than for all of the vertical clouds in Figure 2 (Table 2). Table 3 shows mean and median Seff in Figure 2 within various N1% intervals. The undetermined extreme Seff values beyond the S range of the CCN spectra add uncertainty to these mean Seff but not to the uncertainty of the median Seff, which are probably more representative anyway. The last two rows of Table 3 are another attempt to make more relevant estimates of Seff by using the 90th percentiles of Nc for all of the 1 second measurements within each of the clouds. Because of the reductions of Nc by entrainment and coalescence, the Seff estimates based on mean Nc underestimate cloud S that is due to the initial aerosol influence at cloud base. Thus, the higher Seff in the last two rows of Table 3 may be a closer approximation of the Seff that better reflects the input aerosol influence.

Figure 4.

As Figure 2 but for the clouds shown in Figure 3.

Table 3. Mean and Median Seff Within N1% Rangesa
N1% (cm−3)<100100–200200–300300–400>400
  • a

    First two rows are from Figure 2. Last two rows are from a similar figure using instead of mean Nc the 90th percentiles of Nc measurements of the 1 second intervals within each cloud to match with NCCN at various S.

Mean Seff (mean Nc)0.870.510.460.430.11
Median Seff (mean Nc)0.720.330.370.280.07
Mean Seff (90% Nc)1.330.850.630.580.17
Median Seff (90% Nc)1.380.750.460.340.13

[12] The Seff-NCCN R and slopes of the linear regressions are similar for all S values of the CCN spectra and for Seff-NCN but the peak Seff-NCCN R values are all close to S = 1% (Seff-N1%, Table 2). R for Seff-Nc is also negative (one small exception), but with smaller absolute values (Table 2) than R for Seff-N1%. The less negative R for Seff-Nc than Seff-NCCN is because both Seff and Nc are functions of NCCN.

[13] Figures 2 and 4 and Table 3 display similar progressive decreases of Seff with NCCN for the vertical and horizontal cloud penetrations and the linear LWC clouds. R for Nc-NCCN are also similar for all S of NCCN and maximal for N1% for all of the vertical and horizontal clouds. This is consistent with the high cloud S indicated in Table 3 and Figures 2 and 4.

[14] Along with NCCN, W also determines Nc; i.e., higher W produces higher S and Nc. As expected Nc increased with within-cloud W (positive R for Nc-W in Table 2 except last column), but with much lower R than Nc-N1% (row 2). Variations of W thus contribute to the scatter in Figures 14 because air motion (W) variations ought to be independent of aerosol (i.e., NCCN, N1%) variations. However, N1% and W are positively correlated (row 4; except last column). But since the effects of NCCN and W on Seff are opposite (higher NCCN reduces Seff whereas higher W increases Seff) this positive albeit small coupling of NCCN with W is probably the reason that Seff does not display the expected positive correlation with W (Table 2). Apparently this coupling allowed the greater effect of NCCN variations on Seff to overwhelm the effect of the W variations on Seff. This means that the decrease of Seff with N1% shown in Figures 2 and 4 and Table 3 could have been even greater had W been constant or independent of NCCN. Consistent with this is the fact that for the 17 adiabatic clouds W and N1% were not correlated (R = −0.28; last column of row 4) and this probably contributed to the greater negative R of Seff-N1% for the adiabatic clouds compared to all 69 vertical clouds (0.84 versus 0.74) in spite of the more limited N1% range of the adiabatic clouds, which thus prevented observations of the greatest reductions of Seff caused by the highest N1%. Furthermore, the larger positive R for N1%-W of the 29 LSA vertical clouds than the 40 HA vertical clouds (0.53 versus 0.30) contributed to the smaller negative R of Seff-N1% for the LSA clouds compared to the HA clouds (−0.35 versus −0.86 in Figure 2 and Table 2). But again this lower R for the LSA clouds is also due to the smaller N1% range for these cases, which prevents the most extreme reductions of cloud S due to the highest concentrations. On the other hand, the greater positive R of N1%-W for horizontal than vertical (0.51 versus 0.39) should not have contributed to the smaller magnitude R of Seff-N1% for vertical versus horizontal in Figures 2 and 4 (−0.74 versus −0.83).

4. Conclusions

[15] This study demonstrates the systematic decrease of cloud supersaturations (S) with higher CCN concentrations (NCCN) that has only been demonstrated in stratus for two extremely different air masses [Yum and Hudson, 2002]. The dependence of Seff on NCCN limits the value of reported Seff without specification of the NCCN range of the observations. This study indicates higher cloud S in stratus clouds than has been indicated by most previous studies; i.e., those mentioned in the first sentence of the Introduction. The present results are not in conflict with many of those earlier studies where concentrations were often higher and thus consistent with the lower S at higher NCCN and Nc observed here. Moreover, the high stratus S found here at low NCCN and Nc are essentially consistent with a couple of previous studies in cleaner air masses [Yum et al., 1998; Yum and Hudson, 2002]. The most important result is the consistently high Seff of the cleaner stratus clouds, which approached or exceeded 1% even when based upon average Nc. Seff based on mean Nc is often a marked underestimate of the S that had resulted from the initial influence of CCN on Nc.

[16] Higher stratus cloud S means that a larger subset of atmospheric particles are capable of nucleating stratus cloud droplets; i.e., pure NaCl particles would only need to be 20 nm diameter to activate at 1% S rather than 50 nm for 0.3% S. Pure ammonium sulfate particles would only need to be 28 nm rather than 60 nm. This has important implications for the indirect aerosol effect and the geoengineering proposal to whiten maritime stratus clouds. The higher Seff found here in cleaner maritime stratus would make these abundant clouds easier to manipulate with easier-to-produce smaller seed particles. Further efforts will make better estimates of adiabatic Nc [i.e., Hudson and Yum, 2002] to provide better Seff estimates that will be compared with predictions of Nc and Seff based on CCN spectra and W.

Acknowledgments

[17] This work was supported by the NSF ATM-0734441. Herman Gerber organized POST. CIRPAS pilots, ground crew, and scientists provided an ideal platform for the superbly executed flights and provided the cloud microphysics data.