Atmospheric aerosol particles (APs) modify the planetary albedo, and hence climate, via the following mechanisms: (1) by altering short-wave light scattering and absorption (Charlson et al., 1992), and (2) by acting as cloud condensation nuclei (CCN), thereby altering the radiative properties and lifetime of clouds (Charlson et al., 1987; Albrecht, 1989). These effects of APs on climate, known as the direct and indirect effects respectively, are influenced by the AP size distribution and composition, which are, in turn, determined by AP sources and sinks.
One possible source of condensation nuclei (CN) is new particle formation from gaseous precursor species. Formation of CN produces very little AP mass but a large number of particles, significantly altering the shape of the AP size distribution. Due to the importance of CN formation, many observational (Weber et al., 1999, 2001; Birmili et al., 2000) and numerical (Kulmala et al., 1995; Clarke et al., 1999a; Korhonen et al., 1999) studies have been undertaken to assess the major mechanisms responsible for their generation, the exact details of which are still uncertain.
Previous studies have found high levels of sulphuric acid (H2SO4) in association with CN formation, suggesting that binary homogeneous nucleation of sulphuric acid and water (H2O) and subsequent condensation are the main source of particles in the atmosphere (e.g. Clarke et al., 1999a). However, observations of CN formation rates cannot be reconciled with theoretical models of H2SO4– H2O binary nucleation (Birmili and Wiedensohler, 2000), because the measured H2SO4 concentration is generally insufficient (Weber et al., 1999). Nucleation involving other precursor species, such as ammonia (NH3), has been proposed (Coffman and Hegg, 1995; Korhonen et al., 1999) and shown to increase the rate of CN formation (Ball et al., 1999), especially during enhanced solar radiation (Birmili et al., 2000). Several recent theoretical (Kulmala et al., 2004, 2006a) and laboratory (Kulmala et al., 2006b) studies have proposed that soluble inorganic compounds play a second-order role in new particle formation and that water-insoluble organic vapours contribute significantly to new particle formation.
Another mechanism proposed to enhance CN formation is due to small fluctuations in temperature and relative humidity (RH). Such fluctuations can be generated under favourable atmospheric conditions such as breaking Kelvin– Helmholtz waves (Bigg, 1997) or mixing processes in the atmosphere (Nilsson and Kulmala, 1998). Easter and Peters (1994) showed that a change in temperature of 2– 3 °C or 6– 8% in RH could result in a factor 10 increase in the nucleation rate. Schröder and Ström (1997) observed frequent upper-tropospheric particle production in the vicinity of a cold front, which they attributed to dynamically induced mixing processes. Ion-induced nucleation has also been proposed as a mechanism for new particle formation (Yu and Turco, 2000; Laakso et al., 2002; Lee et al., 2004).
The CN formation rate predicted by the above mechanisms is dependent on several physical parameters: particulate surface area, precursor-gas concentrations, temperature, water vapour pressure and H2SO4 vapour pressure. In the boundary layer (BL), Covert et al. (1992) attributed formation of CN to low (≤ 5 µm2 cm−3) ambient AP surface area. In the free troposphere (FT), Clarke et al. (1998) observed particle production in the outflow of clouds when the AP surface area decreased to about 5– 10 µm2 cm−3. The decreased AP surface area was attributed to cloud processing (rain-out) of the APs, and was considered a necessary prerequisite to new particle formation (Perry and Hobbs, 1994). Twohy et al. (2002) and Clement et al. (2002a) observed extremely high concentrations of new particles in the outflow of a deep convective system, and concluded that cloud outflow regions could be a major source of new particles in the troposphere. Although the decreased AP surface area was attributed as the major parameter determining whether CN formation would occur, increased concentrations of precursor gases lofted from the BL to the upper troposphere were also hypothesised as a requirement for the presence of CN in cloud outflow.
Enhanced actinic fluxes above clouds have been found to induce increased hydroxyl radical (OH) concentration (Mauldin et al., 1997) which will subsequently oxidise with SO2 to form H2SO4, which in turn condenses to form CN. Concomitant with other favourable conditions for CN formation (low surface area, high RH and low temperatures), the enhanced actinic flux above cloud results in favourable conditions for particle nucleation (Weber et al., 2001). Hegg (1991) also observed that the increased enhanced actinic flux within cloud (Madronich, 1987) may be responsible for CN formation in the interior of clouds. Kulmala et al. (2006b), however, showed via theoretical calculations and laboratory studies that the presence of new particles in cloud interiors did not originate from sulphuric acid nucleation but, rather, from insoluble organic vapours.
In this article, we report airborne observations of AP size distributions and CN formation measured during the Asian Aerosol Characterization Experiment (ACE– Asia). The observations are novel because a comparison of two research flights is made: a research flight preceding the cold front of an extratropical cyclone and one behind the cold front. The daily weather patterns throughout the middle and high latitudes are closely related to the passage of extratropical cyclones, consequently they have major implications for the formation, modification and removal processes of the ubiquitous AP populations at these latitudes. The air masses sampled are interpreted in terms of the conveyor-belt model of a front, to identify input and output airflow from the frontal clouds enabling us to determine the effect of a cold front on CN concentrations. Stability and turbulence parameters are examined to investigate possible enhancement of nucleation due to mixing processes, as well as correlations with ambient particle surface area, SO2 and RH. The intentions of this paper are: (1) to present statistics of CN concentrations in the BL and FT of pre- and post-frontal regions, and (2) to examine mechanisms responsible for new particle formation.
The platform used for the Australian contribution to ACE– Asia was the Airborne Research Australia (ARA) Beech B200T Super King Air. The aircraft was equipped with an extensive suite of instrumentation capable of measuring thermodynamics, three-dimensional wind components, cloud and aerosol microphysics, radiation fluxes and trace gases.
CN were sampled with an isokinetic inlet, and counted with separate CN counters able to resolve particles of two different size ranges; a TSI-3025 resolved CN with a radius, rp≥ 1.3 nm and a TSI-3010 resolved CN with a radius, rp≥ 6.0 nm. The minimum detectable radius for the TSI-3025 (1.3 nm) is marginally smaller than the usual minimum quoted in the literature of 1.5 nm. This value was determined after intercomparison of several TSI-3025 CN counters preceding the ACE– Asia experiment (John Gras, personal communication). Larger AP were measured using a Particle Measuring Systems (PMS) Active Scattering Aerosol Spectrometer Probe (ASASP). The ASASP can measure AP in the radius range 0.065– 1.5 µm covered by 15 size bins; it has a heated inlet that dries the AP before size measurement. CN and ASASP measurements presented here are all taken in clear air, where clear air is defined to have a liquid water content (LWC), ql ≤ 10−3 g kg−1 as measured by a Commonwealth Scientific and Industrial Research Organisation (CSIRO) hot-wire probe (King et al., 1978). Exclusion of saturated regions ensures that dry, rather than ambient, AP size distributions are reported (Strapp et al., 1992) and that droplet break-up in cloudy regions did not result in spurious CN counts (Weber et al., 1998). The first and fifth size bins of the ASASP were subject to altitude-dependent detuning problems above approximately 4000 m above sea level (asl). AP spectra sampled above 4000 m asl have been modified by excluding the first bin and linearly interpolating the fifth size bin between the fourth and sixth size bins. Cloud droplets were measured with a PMS Forward Scattering Spectrometer Probe (FSSP-100) capable of sizing and counting droplets in the radius range 1– 23.5 µm in 220 size bins. A PMS 2D-C optical array probe was used to measure droplet spectra in the radius range 37.5– 400 µm in 80 size bins. Temperature was measured with a Rosemount temperature sensor, wet-bulb temperature with a CSIRO wet bulb sensor. Liquid water content (LWC) was measured using the FSSP-100 and a CSIRO hot-wire probe. All thermodynamic measurements and ASASP data are 1 Hz averages of 64 Hz data, while the FSSP and 2D-C measurements are 1 Hz averages of 4 Hz data.
The total concentration measured by the CN counters is referred to by their lower limit of size detection, N1.3 for the TSI-3025 and N6 for the TSI-3010. The concentration difference, N1.3 − N6, is an indicator of nucleation of CN (e.g. O'Dowd et al., 1998) and will be referred to as Ultrafine Condensation Nuclei (UCN) (sometimes also called the ‘nucleation mode’). In addition, another parameter used as an indicator of new particle formation is the ratio UCN/N6 (Warren and Seinfeld, 1985; Covert et al., 1992; Schröder and Ström, 1997; Young et al., 2007).
3. Synoptic overview
3.1. Flight paths
On 24 April 2001, research flights 010424a and 010424b were conducted. Flight 010424a was conducted in the morning/early afternoon in pre-frontal air ahead of a surface cold front, while flight 010424b was flown in cold subsiding post-frontal air during the afternoon/early evening. Flights 010424a and 010424b will hereafter be referred to as the pre-frontal and post-frontal flights, respectively.
A mean sea level pressure (MSLP) map (not shown) and satellite imagery (Figure 1) from 24 April 2001, indicated the presence of an extratropical cyclone to the southwest of Kyushu. The visible spectrum satellite image of the frontal synoptic situation, shown in Figure 1 (obtained using the GMS-5/SVISSR satellite) was obtained at 1324 Japanese Standard Time (JST = UTC + 9 h). The flight tracks for the pre-frontal and post-frontal flights are shown in Figure 2. Both flights were undertaken at about the same latitude and longitude, but advection of the front during the day resulted in the morning flight being flown in the pre-frontal air mass and the afternoon flight in the post-frontal air. The MSLP indicated that the front was advecting northeasterly at approximately 45 km hr−1.
The research components of both flights are projected onto a vertical plane (in the horizontal flight direction) and shown in Figures 3 and 4 for the pre-frontal and post-frontal flights, respectively. Both flights consisted of upper-level legs conducted in the FT, followed by a descent sounding and then legs flown in the BL.
3.2. Vertical structure of the atmosphere
3.2.1. Pre-frontal flight
The King Air completed five legs in the upper troposphere, descending from 8500 m to 6500 m (Figure 3), to examine the upper-level outflow of the frontal clouds. A descent sounding was then made to approximately 100 m asl, during which time a thick (600 m) layer of stratocumulus was encountered, extending from 1400 to 2000 m. Following the sounding, a series of level legs were made below cloud base at 100 m and 500 m asl. At 1 km asl, convective clouds located east of the stratocumulus, were encountered. Finally, a level leg 150 m asl westward behind the frontal boundary was flown. During the penetration of the surface cold front, the forward-looking video revealed a ceasing of precipitation and clearing of the sky with cumulus in the distance, typical of a post-frontal BL (Cotton and Anthes, 1990).
The pre-frontal sounding (Figure 5(a)), obtained during descent, reveals a mixed layer below 1 km asl and stable inversions located at 2 km and 3.8 km asl. The stratocumulus layer was confined below the inversion at 2 km, below which the sounding exhibits a humid structure, typical of warm fronts in the Japan region (Kurihara et al., 2001). The presence of warm stable layers above cold air indicates a warm front located at 2 km asl. Appealing to the conveyor-belt model of a cold front, the upper-level legs and the sounding above 2 km were necessarily conducted in the warm conveyor belt (Houze, 1993). Furthermore, at higher altitudes the air consisted of air originating from the warm and dry conveyor belts. The height of the transition from the BL to the FT in the pre-frontal air mass as determined from the sounding (Figure 5(a)) was also located at about 2 km. At this altitude there was a significant increase in temperature (about 1.5 °C) and change of wind direction, signifying a transition between different conveyor belts of the front. This transition is identified as the transition from the warm/dry conveyor-belt air to the underlying colder air of the cold conveyor belt. The height of the warm front and the direction of flow of the warm and dry conveyor belts are shown in Figure 3.
3.3.2. Post-frontal flight
The afternoon flight was flown in the post-frontal air mass behind the surface cold front. The descent sounding (Figure 5(b)) indicates the height of the BL extended to about 500 m at which point the wind shifts from easterly to near westerly which delineates the transition from the BL to the FT. Also evident is a strong temperature inversion located at about 2.2 km, which separates relatively warm dry air from cold dry air below, indicating where the aircraft penetrated the top of the surface cold front. Since the dry conveyor belt originates in the FT and passes over the surface cold and warm fronts, air above the inversion is most likely within the dry conveyor belt. Due to icing of the wet-bulb temperature sensor, the sounding is unreliable below (above) 600 hPa (4 km). At an altitude of about 4500 m, the aircraft penetrated deep convective clouds generated by convergence at the surface cold front. These clouds were precipitating heavily, typical of narrow cold-frontal rain bands. Approximately 25 km behind the surface cold front, a series of legs were flown within and below some lightly precipitating stratocumulus. The position of the cold front and the direction of motion of the dry conveyor belt are shown in Figure 4.
3.3. Back trajectories
Back trajectories showing the history of air masses during the 72 h prior to the research flights, obtained using HYSPLIT (Draxler and Rolph, 1997), are shown in Figure 6. The meteorological data (Final Global Data Assimilation System run– FNL) are on a horizontal grid of 190 × 190 km and a vertical grid comprising 13 levels spaced from the surface to 20 hPa (Stunder, 1997). Although the resolution of the back trajectories is too coarse to explicitly resolve the cold front, they do reveal information about its air mass history and general conveyor-belt structure.
The trajectories illustrate that air arriving at the sampling location had passed over densely populated areas of Japan and southwestern China. Therefore, the air masses were most likely influenced by anthropogenic emissions. Also the trajectory ending at 3 km agl had undergone substantial uplift during the previous 24 h due to the presence of the front. Upper-tropospheric air advected from northern India, maintaining a relatively constant altitude along its trajectory. Air at 1 km altitude had spent the previous 48 h within 1 km asl suggesting that air at this altitude acquired a marine component, subsequent to a continental influence obtained over Japan. The back trajectories are consistent with the wind directions shown in the soundings (Figure 5), and also illustrate the history of the dry and warm conveyor belts. The trajectory that originates to the east of the end location moves anticyclonically ahead of the approximate position of the surface cold front and descends for most of its history; however, in the final 3 h it undergoes substantial uplift, which indicates that this trajectory is representative of the warm conveyor belt. The two trajectories that start to the west of the end location are located above (at their end times) the height of the surface cold front and, therefore, are contained within the dry conveyor belt.
4.1. Statistical properties of aerosol particles in the FT and BL
Figures 7 and 8 are plots of the observed UCN (N1.3 − N6) concentration with respect to the N6 concentration; each point represents a 1 Hz sample during a horizontal leg in the BL or FT. The statistical properties of N1.3, N6, UCN and UCN/N6, averaged over a flight leg, for the pre-frontal and post-frontal flights are summarized in Tables I and II, respectively.
Table I. Statistical properties of nucleation variables during the pre-frontal flight averaged over each leg as labelled in Figure 3.
231 ± 85
100 ± 33
132 ± 68
1.37 ± 0.70
328 ± 108
155 ± 51
174 ± 81
1.23 ± 0.75
408 ± 154
206 ± 74
203 ± 109
1.06 ± 0.64
230 ± 108
117 ± 63
117 ± 74
1.41 ± 1.20
137 ± 90
71 ± 46
68 ± 58
1.24 ± 1.31
276 ± 109
130 ± 53
139 ± 78
1.26 ± 0.92
1826 ± 784
1344 ± 422
552 ± 576
0.39 ± 0.41
1433 ± 336
1149 ± 240
340 ± 255
0.30 ± 0.23
1033 ± 139
841 ± 64
192 ± 117
0.23 ± 0.14
1717 ± 746
1444 ± 572
344 ± 300
0.23 ± 0.13
1502 ± 501
1195 ± 219
357 ± 312
0.29 ± 0.23
Table II. As Table I, but for the post-frontal flight averaged over each leg as labelled in Figure 4.
940 ± 95
501 ± 50
439 ± 64
0.88 ± 0.12
1034 ± 153
919 ± 124
115 ± 74
0.13 ± 0.10
1995 ± 562
1722 ± 445
272 ± 181
0.16 ± 0.07
1515 ± 358
1320 ± 285
194 ± 128
0.15 ± 0.09
In the pre-frontal BL, N6 concentrations range from about 800 to 3500 cm−3 and average N̄6∼1200 cm−3. The UCN concentration ranges two orders of magnitude from about 10 to 1000 cm−3, with an average N̄6∼360 cm−3. In the pre-frontal FT, N6 concentrations span 10– 300 cm−3 with an average N̄6∼130 cm−3, and UCN concentrations from about 1 to 600 cm−3, with an average of about 140 cm−3. The UCN/N6 ratio was largest in the FT (mean 1.26) and smaller in the BL at 0.29.
In the post-frontal BL, average N6 values were about 1320 cm−3, UCN concentrations averaged about 190 cm−3, and UCN/N6 was 0.15. The UCN concentration and UCN/N6 ratio in the post-frontal BL are about half their respective values in the pre-frontal BL. Substantial differences are also apparent between flights in the FT. In the post-frontal FT, N6 and UCN concentrations show less variability and are about double the values for the pre-frontal flight; the FT UCN/N6 ratio is lower, at approximately 0.9.
The N1.3 and N6 BL concentrations agree within 10% between the pre-frontal and post-frontal air masses, which suggests that air in the warm conveyor belt and in the cold subsiding air behind the cold front have similar chemical sources. This is consistent with the back trajectories (Figure 6), which indicated that BL air, ahead of and behind the surface cold front, most likely contained an anthropogenic influence from Japan. In contrast, the post-frontal FT, which contains air located predominantly in the dry conveyor belt, has marked different AP properties from the pre-frontal FT. The pre-frontal FT contains a mixture of air from the dry and warm conveyor belts; air in the dry conveyor belt has passed over the surface cold front and mixed with air from the warm conveyor belt forced upwards by convergence at the leading edge of the surface cold front. The pre-frontal FT air has therefore been processed by clouds associated with the surface cold front. The post-frontal FT, unperturbed by convection, shows less variability in both N6 and UCN concentrations than the pre-frontal BL, suggesting that mixing of air parcels has created regions both conducive and inhibitive for new particle formation in the FT. The UCN and N6 concentrations are larger in the post-frontal FT which implies that frontal clouds are a net sink for UCN and N6. However, the UCN/N6 ratio is larger in the pre-frontal FT which has two explanations: (1) the pre-frontal FT is more conducive to UCN production than the post-frontal FT, which will increase the numerator in the UCN/N6 ratio or, (2) precipitation scavenging of aerosol particles by the frontal clouds will preferentially scavenge larger particles since it is the larger particles which activate to become cloud droplets, which will decrease the value of N6 in the pre-frontal FT compared to the post-frontal FT. These observations do not preclude that UCN were recently formed near the frontal clouds (Weber et al., 2001), only that scavenging of UCN was greater than their rate of formation.
In summary, the following conclusions are evident from the statistical analysis of nucleation variables:
(1) new particle formation, as measured by the difference of two CN counters operating at different minimum threshold radii, is most prevalent in the post-frontal FT, however the pre-frontal BL is more conducive to new particle formation than the post-frontal BL;
(2) The UCN concentration in the pre-frontal FT was found to be about 140 cm−3 which was substantially smaller than in the post-frontal FT (about 440 cm−3) because UCN have been scavenged by frontal clouds;
(3) The UCN/N6 ratio is larger in the pre-frontal FT (1.26) than in the post-frontal FT (0.15) which could have two explanations: (a) UCN production is greater in the pre-frontal FT as compared with the post-frontal FT or, (b) the frontal clouds have scavenged AP with radius larger than 6 nm preferentially over UCN.
4.2. Horizontal profiles
We now focus on fluctuations in the UCN concentration for each of the horizontal legs. Figure 9 shows measured UCN, RH and temperature, wet equivalent potential temperature, sulphur dioxide and AP surface area (as measured by the ASASP). Wet equivalent potential temperature (θq) was defined as in Pointin (1984), which is an extension of the formulation employed by Paluch (1979) such that it is valid for saturated and sub-saturated air parcels. The following criteria have been been applied to the data: (1) LWC content, as measured by the CSIRO King probe, was less than 1 × 10−3 g kg−1; (2) The ratio UCN/N6 was greater than unity. The first condition is to prevent contributions to particle counts during cloud penetrations. The second criterion has been used in previous studies to identify regions of new particle formation in the upper troposphere (e.g. Schröder and Ström, 1997; Lee et al., 2004; Young et al., 2007). Note that criterion (2) means that samples shown in Figure 9 correspond to all points above the one-to-one line shown in Figure 7. Wet equivalent potential temperature is conserved for reversible adiabatic motions and provides a measure of the combined effects of fluctuations in water vapour content and temperature. The dew-point temperature sensor was subject to freezing during penetration of cloud during Leg 5 which resulted in unreliable humidity measurements. Therefore, temperature, RH and θq are not shown for Leg 5.
Regions of new particle formation, as defined by the above criteria, are evident throughout much of the horizontal extent of all legs. The UCN concentrations are relatively low compared to other studies made in the upper troposphere (e.g. Hermann et al., 2003; Lee et al., 2004; Young et al., 2007), but are still significant in both number concentration and spatial extent, indicating the importance of the upper troposphere as a region conducive to new particle formation. Figure 10 shows the same measured quantities as Figure 9 but for the post-frontal flight. Again, new particle formation is prevalent throughout the horizontal extent of Leg 1, however, consistent with the statistical analysis of section 4.1, UCN concentrations are appreciably larger than measured during the pre-frontal flight.
A comparison between the pre-frontal and post-frontal flights reveals information about the mechanisms that are promoting the formation of new particles. Sulphur dioxide concentrations are larger in the pre-frontal region, indicative of frontal convection lofting precursor gases from the BL to the upper troposphere. Despite larger SO2 concentrations in the pre-frontal FT, UCN concentrations are lower than recorded during the post-frontal flight. Additionally, during the pre-frontal flight, SO2 concentrations exhibit a general increase with decreasing altitude. All of these legs were flown in the outflow of the frontal clouds, so the increase of SO2 with a decrease in height observed during the pre-frontal flight is consistent with air parcels having spent decreasing amounts of time within the frontal rain clouds and undergone less aqueous scavenging. Despite the increase of SO2 concentration with a decrease in altitude during the pre-frontal flight, there is no associated increase in the UCN concentration as the SO2 concentration increases.
The AP surface area, as measured by the ASASP, is shown in Figures 9(e) and 10(e) for the pre-frontal and post-frontal flights respectively. The total surface area is significantly larger during the pre-frontal flight, attaining values approaching 40 µm2 cm−3, whereas the AP surface area is an order of magnitude smaller during the post-frontal flight, remaining below 4 µm2 cm3. This is consistent with vertical transport of AP from the BL to the FT in the frontal clouds. The smaller AP surface area in the post-frontal FT region of the front may be a contributing factor to the elevated UCN concentrations compared to those measured in the pre-frontal FT.
However, it is interesting that during the pre-frontal flight, the observed UCN concentrations do not immediately appear to be inversely correlated with fluctuations in total AP surface area. There are periods, evident during Leg 1 and Leg 4 where the UCN concentration is lower when the AP surface area reaches a maximum, however this is not obvious throughout the complete horizontal extent of each leg, or indeed all legs. For instance in Leg 2, at a horizontal distance of approximately 15– 20 km and again between 30 and 40 km, both UCN and AP surface area attain near-maximum values.
We now explore the possibility that fluctuations in RH and temperature are responsible for much of the observed new particle formation in the upper troposphere. During both flights, fluctuations in temperature of a few degrees K and large gradients in RH of up to 40% are present. The combined contribution of changes in RH and temperature lead to changes of up to 4 K in the traces of θq.
Figure 11 is a mixing diagram of θq versus total water mixing ratio Qtot, for the pre-frontal (Figure 11(a)) and post-frontal flights (Figure 11(b)). These diagrams were first utilised to investigate the entrainment process in cumulus clouds (Paluch, 1979; Jensen et al., 1985; Blyth et al., 1988). Since θq is conserved for reversible adiabatic motions and total water mixing ratio is conserved if there is no precipitation, the mixing diagram can be used to identify the sources of entrained air in cumulus clouds. When used in this capacity, a clear-air sounding in the unperturbed environment near cloud, and the thermodynamic state of cloud base are also required. However, in the present case, since the frontal clouds were precipitating heavily we are unable to use the mixing diagram to evaluate how much of the air detrained from the frontal clouds has originated from the BL. We can, however, use the diagram to provide information about the thermodynamic properties and mixing processes occurring within and between detrainment regions from the frontal clouds.
It is evident from Figure 11 that the thermodynamic characteristics along each horizontal leg (each at a near-constant altitude) exhibit linear mixing when plotted on the mixing diagram. Clement et al. (2002a) observed similar linear mixing patterns in the outflow of a mesoscale convective system over the continental United States, however they used gas phase species (CO2 and NOy) as conserved tracers. The colour coding used for each level leg in Figure 11(a) is identical to that used in Figure 9.
Several conclusions can be drawn from this mixing diagram. In general, there is an increase in water vapour mixing ratio as altitude decreases. The mixing line for Leg 1 shows two distinct mixing lines, one portion that exhibits an increasing water vapour mixing ratio as θq increases, and another portion where water vapour mixing ratio decreases as θq increases. Leg 1 was undertaken at about 8200 m, which is near the height of the thermal tropopause in midlatitude regions. It is therefore probable that the portion of the mixing line that exhibits a decrease of water vapour mixing ratio as θq increases is a consequence of tropopause folding near the extratropical cyclone. Ozone measurements would help validate this observation, however, no ozone measurements were made. Nevertheless, although Leg 1 was flown at a near-constant altitude, air masses with quite distinct thermodynamic properties were encountered. The mixing lines of Leg 2 and Leg 3 are connected, which indicates that, despite the legs being separated by several hundred metres, mixing is occurring. This is a clear indication that the layers are coupled and that vertical mixing is occurring between layers in the outflow of the frontal clouds. The mixing line of Leg 4 is separate from the upper-level legs which indicates that it was undertaken at an altitude unaffected by vertical mixing.
It is possible that much of the variation in UCN concentration during each leg is due to mixing between air masses with different particle concentrations (Clement et al., 2002a). To assess this possibility, we evaluated the minimum and maximum (θq, Q) pair for each mixing line and (nominally) assigned the maximum (θq, Q) value as containing a fraction of air F equal to one. Similarly, the minimum (θq, Q) was given an F fraction equal to zero. In this manner, we can determine the degree of mixing for each (θq, Q) pair along the mixing line by (Jensen et al., 1985)
where the subscripts 0 and 1 correspond to an F-fraction of 0 and 1, or in other words, the minimum and maximum (θq, Q) pair, respectively. Therefore, if the fluctuations in the UCN concentration for each leg can be described solely by mixing of air masses with different particle concentrations, then the UCN concentration, as a function of the F-fraction, will also exhibit a linear relationship. Figure 12 shows plots of the UCN concentration against the F-fraction, as defined in Eq. (1). There is a general linear trend observed during all pre-frontal legs, suggesting that some of the observed fluctuations in UCN concentration were due to mixing between air masses with different initial UCN concentrations. For all of the pre-frontal legs, the trend is towards increasing UCN concentrations with decreasing F-fraction. Recalling that an F-fraction of zero corresponds with the smallest (θq, Q) pair on the mixing diagram (Figure 11), then Figure 12 suggests that most UCN are produced in regions of lowest temperature and water mixing ratio. Theoretical parametrisations of new particle formation predict the nucleation rate to be a nonlinear, and particularly sensitive, function of temperature and RH (Vehkamäki et al., 2002; Spracklen et al., 2005). For instance, the Vehkamäki et al. (2002) parametrisation predicts that a temperature decrease of 5 K, or an RH increase of 20%, will result in an order of magnitude increase in the nucleation rate. Our observations (Figure 12) are consistent with particle nucleation being favoured in regions of lower temperature.
However, there are many departures from linearity indicative of processes other than linear mixing being responsible for the fluctuations in UCN concentration. Of interest is the large change in UCN concentration over a narrow range of F-fraction (F ≈ 0.6) during Leg 1, which suggests that a localised mixing event has resulted in parcels which are more conducive to the formation of new particles. We now identify where, along the extent of Leg 1, this mixing event has occurred. Figure 13 summarises the information from Figures 9, 11 and 12 for Leg 1; the separate mixing lines have been distinguished with different colouring. It is apparent that the large range of UCN concentration over the narrow F-fraction range occurred where the dry intrusion was encountered. This is evidence that mixing of air, with different thermodynamic properties has been responsible for the burst of new particles encountered at a horizontal distance of 60– 70 km during Leg 1. Similar bursts of new particle formation have recently been observed by Young et al. (2007).
In conclusion, new particle formation in clear air, as defined by the criteria that the ratio UCN/N6 was greater than unity, was evident in both the pre-frontal and post-frontal regions. The presence of large AP surface area did not appear as a factor in controlling whether new particle formation would occur, however it did affect the number concentration of UCN; UCN concentrations were larger in the post-frontal region where AP particle surface area was smallest. Despite larger UCN concentrations in the post-frontal FT region (as compared to the pre-frontal FT), SO2 concentrations were lower, suggesting that increased concentrations of SO2 do not necessarily result in increased concentrations of UCN.
Some of the observed fluctuations in UCN concentration along level-flight legs could be explained simply by dilution due to mixing of air parcels with differing UCN concentrations, a result also found by Clement et al. (2002). However, the major factor controlling the formation of UCN was found to be mixing of air parcels with differing thermodynamic characteristics. These results are in accord with the hypothesis of Bigg (1997), the modelling studies of Easter and Peters (1994) and the recent observations of Young et al. (2007).
4.3. Vertical profiles
4.3.1. Pre-frontal flight
Vertical profiles of measurements for the pre-frontal flight are shown in Figure 14. Total particle concentrations N1.3 and N6, are shown in Figures 14(a) and (b) respectively. The profile of UCN and UCN/N6 reveals a plume of newly formed particles, constrained below 500 m asl, which has been highlighted by horizontal shading. The region of new particle formation is located below a cloud layer and slightly stable layer at 1 km asl. Identification of regions of new particle formation by high concentrations of UCN is arbitrary and we have only included the obvious peak in UCN. For example, Weber et al. (2003) considered areas of new particle formation to be regions where the UCN concentration exceeded 100 cm−3 which, if applied here, would extend the region of new particle formation to 1 km asl. Accumulation mode AP, as measured by the ASASP, had a maximum concentration of about 900 cm−3 and surface area 150 µm2 cm−3. The SO2 concentration was nearly constant at 0.5– 0.6 ppbv.
The plume of new particles is in a region with large AP surface area; additionally, the AP surface area is constant throughout the vertical extent of the particle formation event. Model studies (e.g. Clement et al., 2002b) have shown that a large AP surface area will suppress nucleation events, if the SO2 concentration is less than a specific value, due to the competing mechanisms of gas-phase nucleation and coagulation scavenging to pre-existing AP. In a plume with large AP surface area such as this, Weber et al. (2003) indicated that new particle formation would proceed if the SO2 concentration was above about 2 ppbv (the measured SO2 concentration was about 0.5 ppbv). However, recent calculations by Kulmala et al. (2006a) have shown that only a minimal amount of SO2 may be required for the formation of 3 nm (diameter) particles via cluster activation, as opposed to classical binary homogeneous nucleation. Nevertheless, the particles subsequently formed are subject to rapid Brownian diffusion to existing AP, so it appears that whichever nucleation mechanism is responsible, it is sufficient to overcome the quenching due to the large measured AP surface area.
The large AP surface area and concentration between 3000 and 3700 m and between 7500 and 8000 m asl were revealed, by examination of 2D-C images, to be due to the presence of ice particles. Ice impacting on the ASASP shattered, resulting in a spurious increase in concentration and surface area. Between 3000 and 3700 m asl there is also an increase in SO2 concentration from 0.5 to 1.7 ppbv. The back trajectories (Figure 6) indicate that air at this altitude had originated from the BL over mainland China, where it acquired a large anthropogenic SO2 component. Despite the substantial increase in SO2 concentration, there was no corresponding increase in the formation of UCN, indicating that increased levels of SO2 are not necessarily a prerequisite for the nucleation of new particles. It is possible that no new particle formation was observed because the SO2 concentration was below the threshold value (e.g. ∼2 ppbv) required for new particle formation to be detectable.
To examine if spatial gradients and turbulent fluctuations of temperature and RH have played a role in the generation of new particles, stability and turbulence parameters have also been plotted. Figure 14(h) shows the stability of the pre-frontal atmosphere in terms of the virtual potential temperature gradient. If the virtual potential temperature gradient profile is positive, the atmosphere is stable; if negative, the atmosphere is unstable; if zero, the atmosphere is neutral. Figure 14(i) shows the gradient Richardson number Rig, which represents the ratio of turbulence due to buoyancy relative to turbulence due to shear, and is given by
The Richardson number is a measure of whether the flow is turbulent or laminar. Generally, when Rig is larger than the critical Richardson number, Ric, of 0.25, the flow becomes less turbulent and more closely resembles laminar flow (Jacobson, 1999). The region of new particle formation has a RH gradient of 10% (Figure 14(j)), turbulent flow, and a transition from a stable to neutral atmosphere, suggesting that turbulent eddies in the BL are responsible for conditions conducive for new particle formation. Between 7000 and 7700 m asl is another region of enhanced UCN concentration, especially compared to lower altitudes. Again, this region exhibits a gradient of a stable to a neutral atmosphere, large RH fluctuations (> 30%) and turbulent flow.
Hegg et al. (1992) conducted a modelling study of particle production and concluded that mixing of air at the top of the BL down to the surface created conditions favourable for particle nucleation. Rather than appealing to temperature and RH fluctuation on the nucleation rate, he proposed that higher RH aloft resulted in larger particle surface areas than close to the surface. As dry AP sizes have been measured, it is possible that the 10% decrease in RH throughout the nucleation region has lowered the AP surface area such that it falls below a nominal threshold for particle nucleation.
4.3.1. Post-frontal flight
In contrast to the pre-frontal vertical profile, the post-frontal profile contains significant UCN concentrations throughout the extent of the BL and FT (Figure 15). In addition, two regions of enhanced UCN concentration, indicated by horizontal shading, exist between 2000 and 4200 m asl. Of particular interest is the minimum in UCN concentration, located at about 2700 m asl, that separates the maxima in UCN concentrations. The minimum is intriguing as it begs the question, ‘What characteristics differentiate the minimum from the areas of enhanced UCN concentration above and below?’
The largest UCN concentrations are where the ASASP integrated concentration and surface area reach maxima of approximately 500 cm−3 and 80 µm cm−3 respectively. This is a large integrated surface area, but is consistent with the back trajectories (Figure 6) which indicated that air at this altitude originated in the BL over China and acquired a substantial continental and anthropogenic influence. There is a negative correlation of UCN concentration with RH. The correlation coefficients of UCN with ASASP measured surface area and RH are + 0.7 and – 0.7 respectively. The SO2 concentration is nearly constant at approximately 0.25 ppbv throughout the extent of both nucleation regions and has no correlation with UCN concentrations.
The minimum in UCN concentration separating two layers of increased UCN concentration may be due to differing source histories of the layers. Since SO2 is constant throughout all layers, differing sources would implicate the presence of a third species, possibly NH3, as the driving mechanism for nucleation. However back trajectory analysis (Figure 6), showed no significant difference in the history of air masses at 2000, 3000 and 4000 m asl, suggesting that the minimum is due to local phenomena rather than differing parcel histories.
Another possibility for the minimum in UCN concentration between 2600 and 2800 m asl is that the formation of new particles is driven by turbulent fluctuations of temperature and RH. The minimum in UCN concentration is located in a region that is approaching laminar flow, is stable and of a higher RH than layers above and below. The nucleation regions exhibit abrupt transitions from near-laminar to turbulent flow and also stable to neutral thermodynamic profiles. However, at 3400 m asl similar conditions exist in the region of highest UCN production, in direct contrast to the minimum in UCN concentration at 2600– 2800 m asl. If the layers are decoupled and turbulent mixing is confined to mixing within layers of enhanced UCN concentration, and not between the layers, then mixing is a possibility for the enhanced UCN concentrations.
Alternatively, the decrease in nucleation between the nucleation layers (at ≈ 2700 m), may be a consequence of previous convection. The minimum in UCN concentration is just above the inversion at the top of the cold front, which combined with the increase in RH and low ASASP surface area, suggests that it coincided with a moist layer generated by evaporated clouds. However, conditions of low AP surface area and high RH are exactly the conditions previous studies have suggested as being responsible for new particle formation (e.g. Radke and Hobbs, 1991; Perry and Hobbs, 1994; Clarke et al., 1998, 1999a,b). The absence of new particle formation in this region, coupled with the uniformity of SO2 concentration throughout the nucleating and non-nucleating regions implies that a third, water-soluble species (e.g. NH3), may be required for nucleation of new particles.
Finally, due to the close correlation of UCN concentration with AP concentration and surface area and anticorrelation with RH, the UCN may be a product of surface reactions on AP. As the RH decreases, previously absorbed gases can desorb from the AP back to the gas phase and undergo subsequent nucleation. Species likely to undergo such desorption are NH3 and HCl, which can then nucleate via binary (NH3– HCl) or ternary (NH3– HCl– H2) nucleation (Kulmala et al., 1995; Wiedensohler et al., 1997). In the absence of measurements of these gases, this proposition is difficult to substantiate.
There are regions of UCN generation higher in the FT, at approximately 6000 m asl and 7200 m asl. At this altitude the ambient ASASP surface area of 3 µm2 cm−3 is significantly less than in the lower troposphere. The increased UCN concentrations in these layers correlate with the SO2 concentration and enhanced turbulent mixing, again suggestive of the importance of turbulent mixing for new particle formation.
To summarize, the above results substantiate that turbulent mixing and the associated RH and temperature fluctuations enhance the generation of UCN, however, mixing alone may not be a sufficient criterion for UCN generation. The availability of SO2 and another unidentified gaseous species is also required for nucleation to proceed. Importantly, low AP surface areas were not found to be a necessary factor for UCN formation to occur and, in the post-frontal case, actually had a positive correlation coefficient (0.7) with UCN concentrations.
4.4. Mesoscale effects of the front on UCN, AP and SO2
The post-frontal FT contains higher concentrations of UCN than the pre-frontal (Figures 14(c) and 15(c)), however the ratio UCN/N6 is generally higher in the pre-frontal air. For this to occur, UCN must be scavenged proportionally more than N6 during advection through the frontal clouds. As found in section 4.1, this implies that the front acts as a sink for UCN in the FT. Although the frontal clouds act as a net sink for UCN, this does not rule out that new particles were formed either in the frontal outflow or near the frontal clouds, only that scavenging processes outweigh formation processes.
The average ASASP-measured pre-frontal BL concentration was about 370 cm−3, and the post-frontal concentration about 190 cm−3. The effect of the front was to deplete the AP concentration (in the size range of the ASASP). In the FT, however, the ASASP measured AP concentrations were around 10 cm−3 for both flights, meaning that the convergence associated with the surface cold front has not been effective at transporting AP in this size range to the FT. In contrast, the concentration of SO2 in the FT during the pre-frontal flight is double that during the post-frontal flight (Figures 14(g) and 15(g)). Thus, the front, especially the warm conveyor belt, has been effective at transporting SO2 from the BL to the FT. We also note that the concentration of SO2 in the pre-frontal outflow is double the concentration that entered the frontal system in the dry conveyor belt, however the concentration of UCN is (at least) a factor of five less, which is consistent with the requirement of a third reactive species for the formation of UCN. If this third species had been scavenged effectively during its vertical transport through the frontal clouds, then even the enhanced concentrations of SO2 in the FT would not have been sufficient to generate new UCN. A highly water-soluble gas, such as NH3, fits this criterion.
5. Summary and conclusions
Airborne measurements of CN, APs and SO2 were made preceding and following the passage of a cold front, south of Kyushu during the Aerosol Characterization Experiment – Asia (ACE– Asia). Measurements were obtained in the BL and up to 8.5 km in the FT. Two condensation nuclei counters, a TSI-3025 and TSI-3010, with lower particle radius detection limits of 1.3 nm and 6 nm respectively, measured total particle concentrations, N1.3 and N6. The presence of newly formed UCN was inferred from the difference in concentration between the two counters and also the ratio, UCN/N6. AP concentrations were measured with an ASASP probe.
Thermodynamic soundings in the pre-frontal and post-frontal regions enabled the identification of the major air masses according to the Norwegian Conveyor Belt Model of a cold front. This enabled us to identify the air mass input/output from the frontal clouds. In this picture, the pre-frontal region represents the perturbed state of the atmosphere as it contains cloud-processed air, especially in the detrainment regions of the frontal clouds in the FT. The post-frontal region represents the state of the atmosphere unperturbed by the front. Comparisons of UCN, CN, AP and SO2 were then made between the pre-frontal and post-frontal BL and FT regions of the front to investigate how the cold front processed AP and contributed to new particle formation.
Statistical analyses of UCN, CN and AP concentrations and the UCN/N6 ratio enabled the mesoscale effects of the front on these variables to be determined. The front did not significantly perturb CN concentrations in the BL, which was attributed to the pre-frontal and post-frontal air masses having similar, anthropogenically perturbed, histories. In the FT, however, the pre-frontal region contained significantly fewer UCN and CN than the post-frontal region, suggesting that cold fronts are a significant sink of UCN in the FT. In addition, the frontal clouds were effective at transporting SO2 from the BL to the FT, since SO2 concentrations in the detraining regions of the front were nearly double those in the post-frontal FT. Despite the increased concentration of SO2 in the pre-frontal FT compared to the post-frontal FT, UCN concentrations were subtantially larger in the post-frontal FT. These results suggest that either increased SO2 does not necessarily result in increased UCN, or that a third species, soluble in water such that it was scavenged in the pre-frontal clouds (e.g. NH3), was also involved in the formation of UCN.
Mixing diagrams of conserved thermodynamic variables wet equivalent potential temperature and total water mixing ratio were constructed for horizontal legs in the FT. The mixing diagrams revealed that the thermodynamic properties of air in the FT could be described by linear mixing. Additionally, in the detraining outflow of the frontal clouds, the mixing diagrams revealed the occurrence of large-scale vertical mixing, and an intrusion of dry air, which was attributed to tropospheric folding. UCN concentrations were then plotted as a function of the amount of mixing to examine whether fluctuations along a horizontal leg could be explained by mixing alone. It was found that many of the fluctuations in UCN concentration could be explained by mixing of parcels with differing initial UCN concentrations. However, deviations from linearity were found, especially associated with large-scale mixing events, suggesting that mixing of air masses with distinct thermodynamic properties provides a source of new particles in the FT.
Vertical profiles during the pre-frontal and post-frontal flights revealed several areas of new particle formation which enabled further examination of the processes responsible for UCN formation near the front. New particle formation was dominant in the post-frontal FT in areas of highest particle surface areas, lowest RH, and constant SO2 concentrations. In a localised 2 km section of the pre-frontal profile, there was a 0.7 correlation of UCN concentration with AP surface area and a – 0.7 correlation with RH, showing that enhanced particle surface areas do not necessarily inhibit new particle formation. Analysis of stability (dθv/dz) and shear-induced turbulence (Richardson number) parameters inferred that the areas of highest UCN concentrations were located in transitions from stable to neutral vertical profiles and from laminar to turbulent flow. These observations suggest that mixing of air with different temperature and RH and possibly the presence of NH3 was a major factor in the formation of UCN.
In the BL, the concentration of AP measured with an ASASP probe decreased from 370 to 190 cm−3 after the frontal passage. In the pre-frontal FT, the outflow from the frontal clouds had an average total AP concentration, N6 = 130 cm−3, but during the post-frontal flight, N6 measured 500 cm−3. Thus, the frontal clouds reduced accumulation-mode-sized AP concentrations in the BL and FT. The observations indicate that cold fronts provide rapid vertical transport of BL air and SO2 into the FT, however their overall tendency is to remove aerosol from both the BL and FT and, due to their prevalence in midlatitudes, provide an efficient removal mechanism of aerosol from the atmosphere.
The authors would like to thank Paul Krummel (CSIRO) for supplying some thermodynamic analysis modules. The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and READY website (http:// www.arl.noaa.gov/ready.html) used in this publication. We would also like to thank several anonymous reviewers who provided useful comments on earlier versions of this manuscript.