Results Directory UMM :Data Elmu:jurnal:A:Atmospheric Research:Vol54.Issue4.Aug2000:

with aerosol properties in the vicinity of clouds. An examination of possible aerosol- growth mechanisms is also presented.

2. Experimental

The UMIST Cessna 184, modified for atmospheric research studies, was fitted with a Ž . Particle Measuring Systems PMS ASASP-X and FSSP-100 optical particle counters to examine atmospheric aerosol concentrations and size distributions. The ASASP-X measures the accumulation mode aerosol size distribution from 0.05 mm radius up to 1.5 mm. The FSSP was configured for in situ sampling of coarse mode aerosol ranging from 0.25 to 8 mm. To examine aerosol volatility and, thus, infer chemical composition, a thermal analytical volatility system was also installed which comprised four heater tubes set to specific temperatures to volatilise common aerosol species. The four heater tubes were configured to temperatures of 408C, 808C, 1508C and 3408 C. Aerosol measured at 408C is considered a dry aerosol and is therefore independent of relative humidity Ž . changes O’Dowd et al., 1993 . The loss of aerosol between 408C and 808C is thought to correspond to loss of low volatility aerosol such as nitric acid andror light organic aerosol, while the loss between 808C and 1508C corresponds to the loss of sulphuric acid. Ammonium sulphate volatilises at 2008C and therefore is determined by the loss of mass and concentration between 1508C and 3508C. The residual aerosol remaining at Ž 3508C is normally soot carbon in polluted and non-marine air masses Jennings et al., . 1994; Smith and O’Dowd, 1996 . For a more detailed explanation of the volatility Ž . technique, the readers are referred to O’Dowd and Smith 1993 . The aerosol measured by the ASASP-X was sampled from an intake duct channelling air into the fuselage where it could be selected for direct sampling or volatility analysis. Particle losses were calculated to be negligible for direct sampling and comparison between the direct and volatility sampling indicated approximately 10 loss in the volatility system for sizes Ž . larger than 0.5 mm O’Dowd and Smith, 1996 . The observations reported here were taken in the vicinity of Carlisle in the northwest Ž . of England 54.38N 2.78W and the Eden Valley about 200 km south of Carlisle. Meteorological conditions during the two projects, Winter and Summer, were charac- terised by persistent high-pressure systems. During the Winter project, the centre of the high pressure was a few tens of kilometres to the west of our sampling area at the beginning of the week, however, later in the week the centre strengthened and moved towards central Europe. Two high-pressure systems straddled the UK during the duration of the Summer project.

3. Results

3.1. General cloudy and cloud-free aerosol spectral characteristics During a total of 11 flights in the Winter of 1991 and the Summer of 1992, aerosol characteristics were measured in the boundary layer and the free troposphere under different thermodynamic situations. Flights were conducted under cloudy and cloud-free conditions and the variation in size distributions encountered under different meteorolog- Ž . ical conditions is summarised in detail by O’Dowd and Smith 1996 . In summary, they found that the aerosol size distributions on cloudy days were observed to be substan- tially different from those on cloud-free days, and also substantially different to that Ž . observed in the cloud-free free troposphere CFFT . Aerosol size distributions in a Ž . cloudy boundary layer CBL were typically biased towards larger sizes when compared Ž . with those in a cloud-free boundary layer CFBL , and thus, possessed greater mass. Accumulation mode mode-radii on cloudy days were typically 0.1 mm or greater compared with a mode-radius of f 0.05 mm or less under cloud-free conditions. The larger mode-radii associated with the accumulation mode aerosol when clouds are present suggest that the clouds may provide an effective growth medium to produce particles of this size. Furthermore, in a CBL, aerosol sampled in the vicinity of clouds exhibited greater number and mass concentrations compared with the general back- ground aerosol, further suggesting that the presence of clouds can enhance particulate mass. However, when air, which had recently encountered clouds had advected over strong pollution sources, the minimum in the distribution at 0.05 mm disappeared. In this study, we focus on the difference between background rural aerosol spectral differences under conditions when clouds were present and when no clouds were present. The difference between cloudy and cloud-free size distributions observed by O’Dowd Ž . and Smith 1996 can be best illustrated in Fig. 1a using three typical examples which represent general characteristics observed under these different conditions. Fig. 1a Ž . displays aerosol size spectra observed in a cloudy well-mixed boundary layer CBL on 06r12r1991, in a CFBL on 10r12r1992, and in the CFFT on 06r12r1991. The size distributions are normalised to 100 cm y3 for ease of spectral comparison. Actual particle concentrations for these cases were: CFBL-N s 404 cm y3 ; CBL-N s 290 cm y3 ; and CFFT-N s 25 cm y3 . The cloudy case was chosen because it represented the Ž ths . only case of solid 8r8 and persistent stratocumulus cloud cover and a strong boundary layer inversion leading to minimal exchange with the free troposphere while Ž . Ž . y3 Fig. 1. a Number and b volume size distributions, normalised to N s100 cm , observed in a CBL, CFBL, and in the CFFT. the CFBL case was chosen because it had the least previous contact with clouds. The CFFT case is chosen simply as a typical example of CFFT spectral shape as seen in all Ž . flights during the campaigns O’Dowd and Smith, 1996 . Both in the free troposphere, where clouds are rarely observed under subsiding conditions, and in the cloud-free boundary, aerosol layer spectra exhibit remarkable similarities with the peak in aerosol concentration occurring at sizes less than 0.08 mm. Although the measurements of the size distributions were limited by the 0.05 mm cut-off of the instrument, the data suggest a size distribution mode-radius of around 0.05 mm or less. The spectral shape of the CFFT size distribution is very similar to that observed by Ž . Hoppel et al. 1994 who reported mono-modal size distributions above cloud tops in both continental and marine air in which marine stratocumulus was formed. By comparison, the aerosol size distribution observed in the CBL possess a much higher mode-radius, typically greater than 0.1 mm. This typical CBL distribution is also Ž . very similar in spectral shape to that observed by Hoppel et al. 1994 who consistently observed an accumulation mode with mode-radius of approximately 0.1 mm concomi- tant with a local minimum in the distribution at approximately 0.05 mm. The recurrent difference between the cloudy and cloud-free aerosol spectra suggest that the high mode-radius associated with the aerosol distribution on cloudy days results from cloud-induced aerosol growth. Accumulation mode aerosol typically has a lifetime of a few days, therefore, it is likely that along with newly formed aerosols which had not encountered cloud, the CFBL comprises some residual fraction of aerosol which has already undergone some degree of previous cloud processing. This is suggested by the relative ‘bump’ in the CFBL distribution at sizes greater than 0.1 mm when compared with the CFFT distribution. Volatility analysis indicate that this ‘bump’ is not due to surface-derived crustal material and is volatile at temperatures characteristic of nss-sulphate aerosol. Ž . When the volumetric normalised spectra are examined Fig. 1b , it is clear that the CBL distribution possesses a much greater relative volume with a volumetric mode-radius of f 0.14 mm and a normalised distribution peak of f 2.5 mm 3 cm y3 compared with f 0.1 mm and f 0.7 mm 3 cm y3 , respectively for the free tropospheric aerosol. The CFBL aerosol possesses volumetric spectral features characteristic of both CBL and free troposphere aerosol, supporting the suggestion that a small fraction by number, but a significant fraction by mass, had previously encountered clouds. 3.2. Specific case study During the Summer campaign, on 24th June 1992, aerosol measurements were taken in a decoupled boundary layer with 5–8r8 ths stratocumulus cover at f 1200 m and 2–7r8 ths cumulus between 500 and 1200 m. Vertical profiles of aerosol concentrations, shown in Fig. 2, illustrate some variability in the concentrations, with average concentra- y3 y3 Ž tions of 236 cm in the surface layer and of 262 cm in the sub-cloud layer the layer . between the decoupling inversion and stratocumulus cloud base . Peak concentrations of greater than 300 cm y3 were observed in the vicinity of cumulus cloud edges. Although the number concentration difference between the two layers is only f 10, mass Ž . Fig. 2. Aerosol, EPT and total water content Q vertical profile for 24r06r1992. Stratocumulus cloud location is illustrated by hatched region between 1300 and 1500 m. Typical cumulus cloud location is illustrated by the black region between 500 and 11 m. concentration differs by f 25, with 2.00 mg m y3 occurring in the surface layer compared with 2.43 mg m y3 occurring in the sub-cloud layer. Decoupling of a mixed boundary layer is normally caused by entrainment of dry free tropospheric air induced by increased localised turbulence generated by cloud top radiative cooling. This process results in limited mixing within the cloud and sub-cloud layer but does not extend down to the surface, and consequently, an equivalent potential Ž . temperature EPT inversion results in the mixed boundary layer, thus decoupling the surface and cloud layer. The inversion separating the surface and cloud layer serves to confine the surface flux of vapour and heat within the surface layer. With the build up of heat and moisture in this layer, some parcels become conditionally unstable leading to cumulus cloud formation. If these clouds have sufficient buoyancy, they can penetrate the surface layer inversion and transport moisture and aerosolrCCN into the decoupled cloud and sub-cloud layer. On this basis, the out-flow air from cumuli, which have penetrated the surface layer inversion, will contain aerosol, sucked in from the surface layer, which has been processed in the cloud. Representative aerosol size distributions taken below cumulus cloud base, around the Ž . edge of a cumulus cloud, and just below stratocumulus cloud base shown in Fig. 3 display a significant and progressive enhancement in number andror mass concentra- tion. The distribution observed at the cumulus cloud edge exhibits particle concentra- Ž y3 . Ž y3 . tions N s 274 cm larger than that observed at cumulus base N s 235 cm . The largest concentration increase is observed at sizes around 0.08 mm radius. The spectral changes are consistent with activation of CCN at sizes smaller than r s 0.05 mm followed by subsequent growth into sizes larger than r s 0.05 mm. The effect of the growth process on aerosol mass is to increase the accumulation mode mass from 1.91 mg m y3 at cumulus base to 2.28 mg m y3 at cumulus cloud edge. The activation of these sub-accumulation mode aerosol, combined with the growth of existing accumulation mode aerosol, leads to an increase in both number and mass concentration in the accumulation mode. Fig. 3. Typical spectra observed under cumulus base, at cumulus cloud edge, and under stratocumulus cloud base. The spectra taken at cumulus cloud edge and stratocumulus cloud base also exhibit spectral changes in the form of a shift of size distribution to larger sizes and a concomitant narrowing of the size distribution. Although the sub-cloud layer average concentration below the stratocumulus cloud deck was N s 262 cm y3 , the distribution Ž y3 . shown comprises a concentration N s 245 cm comparable to that observed in the surface layer in order to illustrate the spectral differences between the two aerosol layers. The change in spectral shape between under cumulus base and stratocumulus base results in an increase of mass from 1.91 to 2.54 mg m y3 . By examining the change in aerosol volatility between the surface layer and the decoupled mixed layer, we can attempt to elucidate the aerosol chemical species which are enhanced during these cloud processing events. Fig. 4 shows the volatility size distributions taken in the surface layer and in the decoupled sub-cloud layer. Each spectrum corresponds to a 10-min horizontal reciprocal run average. The loss of aerosol between 408C and 808C is thought to correspond to the loss of low volatility aerosol such as nitric acid andror light organic aerosol while the loss between 808C and 1508C normally corresponds to the loss of sulphuric acid. Ammonium sulphate volatilises at 1808C and therefore is volatilised between 1508C and 3508C. The residual aerosol remaining at 3508C, possessing a small mode-radius, is normally soot carbon in such Ž . polluted and non-marine air masses Jennings et al., 1994; Smith and O’Dowd, 1996 . A similar spectral change to that observed between the two layers shown in Fig. 1 is also seen for a much longer sample period during the volatility runs. It can be seen from the volatility distributions in Fig. 4 that the non-volatile soot carbon aerosol exhibits almost identical spectra and concentration characteristics in both the surface layer and in the sub-cloud layer. Similarly, the ammonium sulphaterbisulphate distributions exhibit broadly similar concentrations and characteristics. The greatest difference between the two layers occurs for the lowest temperature distributions, which point to an enhance- ment in mass of sulphuric acid and other, more volatile, material. The relative mass Ž . Ž . Fig. 4. Volatility distributions observed in surface layer left and sub-cloud layer right . Ambient dry distributions are sampled at 408C. Low volatility aerosol is volatilised between 408C and 808C; sulphuric acid is volatilised between 808C and 1508C; and ammonium sulphaterbisulphate is volatilised between 1508C and 3008C. Aerosol remaining at 3008C is typically soot carbon. Ž concentration of the low volatility material thought to be a nitric acid andror organic . component increased from 2 in the surface layer to 17 in the sub-cloud layer while the sulphuric acid contribution increased from 20 in the surface layer to 38 in the sub-cloud layer. It should be noted that the volatility runs were conducted approximately 1.5–2.0 h after the profile shown in Fig. 2 was completed and that the difference in average mass loadings between the two layers had increased from 2.09 mg m y3 in the surface layer to 4.0 mg m y3 in the sub-cloud layer. The volatility analysis of the cloud-processed aerosol suggests that sulphuric acid is the primary aerosol species added to, or formed in, the existing aerosol. The analysis also points to the growth of pre-existing aerosol by either nitrate andror organic mass addition. A small enhance- Ž . ment of partially neutralised sulphate aerosol ammonium bisulphate is also observed. These observations are consistent with measurements of chemical growth of aerosol in a cap cloud which is frequently encountered at the nearby ground based cloud station Ž . at Great Dun Fell, where Choularton et al. 1996 also observed the growth of nuclei smaller that 0.05 mm into the accumulation mode after passage through a cap cloud and attributed this growth to aqueous phase formation of sulphate and nitrate in cloud Ž . droplets. Although the relative growth observed by Choularton et al. 1996 is similar to that observed here, their results indicate either a conserved or reduced accumulation mode number concentration. Generally, at this site, there appears to be a reduction in accumulation mode number concomitant with an increase in mean particle mass. One possible explanation for this reduction is that the hilltop cloud sampling station causes enhanced turbulent deposition of cloud droplets to the hill, thereby reducing the number of evaporating aerosol on exit from the cloud. A reduction in accumulation mode nuclei after evaporation from cloud was not observed during the measurements presented here, possibly due to the fact that measurements were taken non-intrusively from an airborne platform. Under these moderately clean rural conditions, the measurements reported here indicate that the majority of the accumulation mode aerosol are effective CCN, as growth is observed in all size ranges. This growth compares well with the growth of Ž . aerosol in moderately clean cumulus observed by Leaitch 1996 where a shift to larger Ž . sizes in the accumulation mode was observed. Further, Leaitch 1996 also observed an increase in number concentration of evaporated nuclei at sizes - 0.08 mm, similar to that observed here after passage through a cumulus cloud. This increase tended to preserve the general characteristics of the distribution by compensating for the growth of pre-existing accumulation mode nuclei, even though the mass of the distribution had increased. The explanation for this observation is that under the higher supersaturations encountered in cumulus clouds, Aitken mode nuclei readily are activated into the accumulation mode, thereby increasing the concentration in this mode. Ž . Leaitch 1996 also reported aerosol characteristics after passage through stratocumu- lus under much more polluted conditions than encountered here and illustrated the presence of a bi-modal accumulation mode. Comparison of interstitial aerosol with that observed outside of cloud suggests that the minimum size of aerosol activated under these conditions was significantly higher than that observed under their cumulus case study and the study presented here. No bi-modal accumulation mode was observed in these measurements, however, the radius of the single mode was observed to increase in size. The complete growth of the mode presented here suggests that almost all of the accumulation mode nuclei were activated into cloud droplets compared to the Leaitch Ž . 1996 case where only a moderate fraction were activated. The differences between these stratocumulus cases are likely to be due to the much more polluted environment in the Leaitch case, and thus, higher aerosol loadings, which will tend to reduce the peak supersaturation reached in cloud as more nuclei compete for the same liquid water. This reduction will increase the size of the smallest activated size along with lowering the Ž . fraction of accumulation mode nuclei being activated. Hallberg et al. 1994 also saw a reduced fraction of aerosol activated in the more polluted cases during the Kleiner Feldberg cloud experiment, along with an increase in the 50 partitioning fraction from f 0.05 to f 0.12 mm, when compared to cleaner aerosol conditions.

4. Discussion