2.1 Measurement site

During CCOPE, a CPC (model 3010; TSI Inc., 2000a) and an ultra high sensitivity aerosol spectrometer (UHSAS) (DMT, 2013) were operated at a residence (37.25^∘ S, 73.34^∘ W, 55 m above mean sea level, m.s.l.) near Arauco, Chile (population 35 000). Arauco is a coastal town on the central Chilean Pacific coast. Our measurement site, hereafter the Arauco site (Fig. 1), was selected because of our aim to characterize aerosols advecting onto South America from the southeastern Pacific. Related to this, our measurements were coordinated with investigations of rainfall inside the domain portrayed in Fig. 1. This study region lies in the South Pacific winter storm track, and rainfall here can be strongly enhanced by the Nahuelbuta Range (Garreaud et al., 2016; Massmann et al., 2017). During CCOPE, several rainfall events were studied using profiling radars and a precipitation disdrometer deployed at Curanilahue (Fig. 1), as well as a network of precipitation gauges. The Arauco site is located on a forested hill; most of the population of Arauco lives east of the Arauco site at an elevation less than 20 m m.s.l.

Figure 1Central Chilean coastal region and the location of the Arauco site, where aerosol measurements were made during CCOPE. Altitude thresholds for the digital elevation map are at 0, 50, 250, 500, 750, and 1000 m m.s.l.

Salient characteristics of the CPC and UHSAS are provided in Table 1. These instruments were operated inside the residence at the Arauco site. In addition, a 3 m meteorological tower was deployed adjacent to the residence. Thermodynamic state (i.e., T, P, and humidity) and horizontal wind speed and direction were measured on the tower. CPC and meteorological measurements (minus wind direction) were acquired from 29 May to 14 August (Table 1), UHSAS measurements were acquired from 29 May to 28 June (Table 1), and wind direction measurements were acquired from 19 June to 14 August.

Table 1Aerosol instruments.

Download Print Version | Download XLSX

2.2 Instrumentation

Here we discuss characteristics of the CPC and UHSAS, sampling of the ambient CCOPE aerosol, data acquisition of CPC and UHSAS measurements during CCOPE, and use of the recorded UHSAS histograms to calculate ASDs. Additional information about the UHSAS is provided in Appendix A. In that appendix we discuss how we validated, in a laboratory, the UHSAS's determination of test aerosol concentration and particle size. During those validation studies we intentionally dried the test aerosols to a relative humidity (RH) ≤15 %. Consequently, the effect of aerosol-bound water on either the physical size or the refractive index of the test particles was negligible. UHSAS sizing of partially dried haze droplets (RH ≤ 60 %), sampled from the ambient atmosphere during CCOPE, and an associated particle size overestimate, is also discussed in Appendix A. In Appendix A, we estimate the particle size overestimate to be ∼20 %.

During CCOPE, the CPC and UHSAS sampled ambient aerosol through a section of copper tube (length = 3 m; inner diameter = 0.003 m; volumetric flow rate = 34 cm³ s⁻¹). The inlet end of the tube (hereafter, the sample tube) was secured below an eave on the west side of the residence at the Arauco site. The Reynolds number (Re) of the flow within the sample tube was 960 and thus well below the value (Re=2300) where laminar flow changes to turbulent flow. Particle transmission efficiencies were evaluated using Eq. (7.29) in Hinds (1999). These are 78 % for D=0.01 µm particles and ≥99 % for D=0.1 µm and D=1 µm particles.

The CPC counts particles larger than D=0.012 µm (Table 1) ¹ up to a maximum concentration of 10 000 cm⁻³. CPC data were recorded once per second (Table 1). The UHSAS measures scattering produced when aerosol particles are drawn through light emitted by a solid-state laser (λ=1.05 µm). By reference to a calibration table (Cai et al., 2008, 2013), the UHSAS microprocessor converts scattered light intensity to particle size and accumulates the derived sizes in a 99-channel histogram. Channel widths are logarithmically uniform (Δlog ₁₀D=0.013) over the instrument's full range ($\mathrm{0.055}<D<\mathrm{1.0}$ µm).

Equation (1) was used to calculate the ASD.

Here Δn_i is the “ith” component of the count histogram and $\dot{V}$ is the aerosol flow rate. During CCOPE, the UHSAS aerosol flow rate and the particle count histogram were recorded once every 10 s (Table 1), and hence the sample interval (Δt in Eq. 1) is 10 s.

3.1 Air mass classification and air parcel trajectories

Locations close to the Arauco site are shown in Fig. 1. A significant pollution source in the region is the Arauco paper mill, which releases 600 t yr⁻¹ of SO₂ (Arauco Woodpulp, 2010). When winds had an easterly component, the paper mill may have affected air quality at the Arauco site. Other pollution sources are Concepción (population 950 000), Coronel (population 110 000), Curanilahue (population 32 000), Lebu (population 24 000), and Cañete (population 32 000). In addition, many residences in the region, including the residence where we operated the CPC and UHSAS, burn wood for residential heating.

In a subsequent section, we compare CPC data from the Arauco site to values measured at NOAA's Trinidad Head (THD) observatory in northern California (41.05^∘ N, 124.2^∘ W, 107 m m.s.l.). The THD dataset includes contamination from local sources (e.g., campfires lit by day visitors at the Trinidad State Beach Picnic Ground). Additionally, McKinleyville, CA (population 15 000), and Arcata, CA (population 18 000), are the two coastal population centers reasonably close to THD. Both are southeast of THD, at distances between 15 and 25 km. Northern California's large population centers (the San Francisco Bay Area and Sacramento) are ∼300 km southeast of THD. An important distinction between the sampling at THD and Arauco is the above ground level (a.g.l.) height of the aerosol inlets. This is 10 and 2 m a.g.l. at THD and Arauco, respectively. We cannot state with any certainty if the lower-height sampling at Arauco made those measurements unrepresentative.

Wind measurements made at the Arauco site (Sect. 2.1) and THD were used to conditionally sample the CPC measurements. At Arauco, wind directions from 180 to 330^∘ were chosen as the clean sector. At THD, the clean sector was chosen from 210 to 360^∘. The clean sectors at Arauco and THD are shown in Fig. 2. Three factors contributed to our selection of the clean sectors: (1) inclusion of winds from either true south (Arauco site) or true north (THD), (2) the same range of angles (150^∘) at both sites, and (3) exclusion of wind from the directions of regional population centers.

Figure 2Clean sector chosen for Arauco (a, 180 to 330^∘) and the clean sector chosen for THD (b, 210 to 360^∘).

Additionally, we used HYSPLIT back trajectories (NOAA, 2016) to conditionally sample Arauco site aerosol measurements associated with onshore-moving air. The back trajectories were initialized at 00:00, 06:00, 12:00, and 18:00 UTC. In addition to these static arrival times, trajectories were calculated with the coordinates of the Arauco site² and with wind fields from the Global Data Assimilation System. The spatial resolution of the wind data is 0.5^∘. Position along a trajectory was evaluated hourly. Trajectories that were over the ocean continuously for 3 d before landfall, and had a direction within the clean sector 1 h before arriving at Arauco, were classified as “onshore” trajectories. There are 20 onshore trajectories that overlap with the availability of CCOPE UHSAS measurements.

In subsequent sections, a set of twenty 2 h data segments, centered on the onshore trajectory arrival times, are further analyzed. Appendix B describes the numerical filter we used to derive the aerosol properties analyzed in Sect. 4.2, 4.3, 4.4, and 4.5. The filter attenuates aerosol property variability occurring on timescales shorter than 100 s. We developed the filter to remove narrow “spikes” in the concentration sequences (CPC and UHSAS), which seem to have originated from local sources of aerosol pollution. The Supplement has plots of filtered aerosol properties corresponding to each of the twenty 2 h segments. Four of these were impacted aerosol variability at scales larger than 100 s. In general, these features were not attenuated by the numerical filter. In these instances we discarded (subjectively) portions of the 2 h segment and retained a subset for the analyses conducted in Sect. 4.3, 4.4 and 4.5.

Trajectory altitude is important for determining the presence of SSA particles. Onshore trajectories originating from relatively close to the sea surface, and thus classified as onshore “sea surface” trajectories, were required to have pressures >980 hPa over their 3 d advection to the Arauco site. A total of 18 of the 20 onshore trajectories were also sea surface trajectories. An example of a sea surface trajectory is shown in Fig. 3a–b. The sea surface wind speed (U), analyzed in Sect. 4.5, is the average of the 6-hourly trajectory speeds in the 6 h window ending 6 h before the trajectory arrived at the Arauco site. The averaging interval is shown in Fig. 3b. Two onshore trajectories, classified as “aloft”, had pressures substantially smaller than 980 hPa over their 3 d advection to the Arauco site.

Figure 3(a) One of the 18 sea surface trajectories that arrived at the Arauco site between 29 May to 28 June; this trajectory arrival occurred at 00:00 UTC 9 June. Black dots are hourly output of the HYSPLIT model; however, for clarity, only every other 1 h point is plotted. (b) Hourly parcel mean sea level (MSL) altitude vs. time; however, for clarity, only every other 1 h point is plotted. The averaged sea surface wind speed (U) was evaluated over the 12:00 to 18:00 UTC interval shown in gray. The MSL altitude was calculated using the pressure output by HYSPLIT (parcel barometric pressure) and the ICAO equation for the Standard Atmosphere (ICAO, 1993). The MSL altitude increases if a larger sea-level is pressure applied in the ICAO equation. This sensitivity is ∼8 m hPa⁻¹.

3.2 Sea salt aerosol

Correlated values of SSA concentration and sea surface wind speed are reported in many publications. In a review of the topic, Lewis and Schwartz (2004; hereafter LS04) used a particle's deliquesced wet size, evaluated at 80 % relative humidity, to group SSA particles into three size classes. In field studies conducted at a coastal site, Clarke et al. (2003) demonstrated that particles sizing in the middle of LS04's small particle size class – those with a dry diameter larger than 0.5 µm or a RH = 80 % wet diameter larger than 1 µm – had a composition that was dominated by sea salt (NaCl).

By restricting our focus to segments of the CCOPE data associated with sea surface trajectories (Sect. 3.1), we will analyze UHSAS measurements of particles with D>0.5 µm (N_>0.5) and will assume that this subset of the ASD corresponds to SSA particles. This lower-limit size is a factor of 2 smaller than the RH = 80 % diameter corresponding to the middle of LS04's small SSA class. This is because we assumed that particle size decreased as the aerosol stream warmed from its ambient temperature to the temperature of the UHSAS measurement. Support for this assumption is provided in Appendix A.

3.3 Moments of the aerosol size distribution

In our analysis, we calculated three moments of the UHSAS-measured ASDs. These are the aerosol concentration (N_UHSAS), aerosol surface area (S_UHSAS), and aerosol volume (V_UHSAS). We symbolize these moments as integrals (Eqs. 2–4).

In these formulae the group $(\mathrm{d}N/\mathrm{d}{\log }_{\mathrm{10}}D)\cdot \mathrm{d}{\log }_{\mathrm{10}}D$ represents the concentration of aerosol particles with a diameter between log ₁₀D and log ₁₀D+dlog ₁₀D. Hence, when plotted versus the logarithm of particle diameter, the area under the dN∕dlog ₁₀D curve is proportional to the size-integrated concentration. This is demonstrated in Fig. 4a–b, where the size-integrated concentration is ∼300 cm⁻³ in onshore-moving air (Fig. 4a), and the concentration is approximately 4 times larger (∼1100 cm⁻³) in air thought to be contaminated by continental sources (Fig. 4b). Also apparent is the right tail of an Aitken mode, at ∼0.06 µm in Fig. 4a (onshore-moving air), the absence of an Aitken mode in Fig. 4b (continental air), at least at diameters detectable by the UHSAS (D>0.055 µm; Table 1), and the presence of an accumulation mode at ∼0.1 µm in both air masses (Fig. 4a–b). Two aspects of these results, i.e., the absence of an Aitken mode plus the dominance of an accumulation mode in polluted coastal air is consistent with ASDs reported in Raes et al. (1997) and in Dall'Osto et al. (2010).

Figure 4Consecutive ASDs recorded by the UHSAS at the Arauco site. (a) ASDs with a relatively small concentration (∼300 cm⁻³), a right tail of an Aitken mode (at ∼0.06 µm), and an accumulation mode (at ∼0.1 µm), in onshore-moving air on 5 June 2015. (b) ASDs with a proportionately larger concentration (∼1100 cm⁻³), an accumulation mode (at ∼0.1 µm), and no evidence of an Aitken mode, in air thought to be contaminated by continental sources (4 June 2015). Time is written in UTC in each panel.

Download

4.1 Comparison of CPC data from the Arauco site and THD

In this section, CPC-measured concentrations from the Arauco site and from NOAA's THD observatory are compared. At THD, CPC measurements were made using a TSI 3760 condensation particle counter. The minimum particle diameter detected by the TSI 3760 (D=0.015 µm; Wiedensohler et al., 1997) is slightly larger than that in the TSI 3010 (D=0.012 µm; Table 1). We ignored this distinction.

The THD dataset spans the years 2002 to 2014. Because CCOPE was a wintertime field study, only December, January, and February THD data are used in the comparison. There are 24 346 data points (hourly averaged) from THD and 5541 classify as clean sector. In comparison, there are 745 data points from the Arauco site (hourly averaged) and 194 classify as clean sector. For both sites, we required a clean sector wind speed >1.5 m s⁻¹ in addition to the clean sector directional criteria (Fig. 2). Because the numerical filter (Sect. 3.1) requires 1 Hz CPC measurements and since 1 Hz measurements are unavailable in the THD data archive, the filter was not applied to either of the data sets analyzed in this section.

Table 2Classification of air mass type.

^* Values rounded to one significant digit.

Download Print Version | Download XLSX

In the following paragraph we compare hourly averaged CPC-measured concentrations from the Arauco site and THD. Because the number of data points in these data sets is different, a particular statistical comparison methodology was applied. The approach followed here compares the Arauco and THD average concentrations by applying the Student's t distribution method (t test), explained in Havlicek and Crain (1988; their Eqs. 10.6 and 10.7). The statistical hypotheses are (A) null hypothesis, averages are equal, and (B) alternate hypothesis, averages are different. We also applied the nonparametric Wilcoxon rank-sum test (rs_test; Interactive Data Language, Harris Geospatial Solutions, Inc.). Statistical inferences that we derive based on the Wilcoxon rank-sum test (not shown) are consistent with what we describe below using the t test.

Two aspects of the Arauco–THD comparison are presented here; more detail is available in Fults (2016). First, clean sector measurements are compared. The mean N_CPC at Arauco is 2759 cm⁻³ (standard deviation σ=1827 cm⁻³). The mean and σ at THD are 858±729 cm⁻³. Figure 5 shows the Arauco and THD N_CPC probability distribution functions. Of note is the larger mode concentration and broader distribution at Arauco. Based on our t test comparison, the Arauco average is larger than the THD average (p<0.01). Second, Arauco and THD concentrations are compared without regard to wind direction. The average at the Arauco site is 2971 ± 1802 cm⁻³, while at THD the average is 1059 ± 855 cm⁻³. These averages are also statistically different (p<0.01), and, again, the Arauco average is larger than that at THD. Based on averages presented in this section and information provided in Table 2, two summary statements are warranted: (1) during wintertime, THD is classified as a moderately polluted marine site, and the Arauco site classifies between moderately polluted marine and heavily polluted marine. (2) These sites are not representative of conditions well removed from anthropogenic influence.

Figure 5CPC concentration probability distribution functions for the Arauco site and THD.

Download

4.2 Continental contamination

In this section we probe why aerosol properties varied strongly during 4 of the 20 onshore trajectories. Among these, the example presented in Fig. 6a–c exhibits the largest degree of CPC and UHSAS variability. During this 2 h data segment, centered on 00:00 UTC 9 June (21:00 local time), winds were light at Arauco and Curanilahue (≤1 m s⁻¹) and the wind direction was variable at Curanilahue (Arauco site wind direction measurements are only available after 19 June 2015; Sect. 2.1).

Figure 6Aerosol properties centered on 1 of the 20 onshore trajectories that arrived at the Arauco site between 29 May to 28 June. This trajectory arrival occurred at 00:00 UTC on 9 June. (a) UHSAS concentration; (b) UHSAS aerosol volume; (c) UHSAS aerosol surface area. Aerosol properties shown here were filtered using the procedure described in Appendix B. Vertical dashed lines mark the subset of the 2 h segment we picked (subjectively) as being representative of onshore-moving air that was relatively unaffected by continental aerosols compared to adjacent portions of the 2 h segment.

Download

Over the ocean, 12 to 6 h prior to 00:00 UTC 9 June, the HYSPLIT wind speed was 8.3 m s⁻¹ and the HYSPLIT direction was westerly (Fig. 3a). In terms of UHSAS measurements (Fig. 6a–c), an obvious feature is the variability in the sequences of N_UHSAS, V_UHSAS, and S_UHSAS. The S_UHSAS is largest during an enhancement at ∼ 00:37 UTC. The question arises of whether winds over the ocean and the resultant SSA production can cause this variability or if continental aerosol sources have to be evoked to explain this phenomenon. This was addressed by calculating aerosol surface areas as a function of wind speeds that bracket the HYSPLIT-derived wind speed (8.3 m s⁻¹). The basis for this calculation is the S-on-U parameterization described in LS04 (their Fig. 22). The calculation indicates that S can range between 6 µm² cm⁻³ (U=6.3 m s⁻¹) and 15 µm² cm⁻³ (U=10.3 m s⁻¹). Since the upper-limit of the predicted variation is small compared to S_UHSAS at ∼ 00:37 UTC (Fig. 6c) and at other times in Fig. 6c and because the wind speed variation applied in the calculation is an order of magnitude larger than the variation in the HYSPLIT-derived wind speed (±0.1 m s⁻¹), it is concluded that the aerosol enhancements seen in Fig. 6a–c are not due to a wind speed increase over the ocean. Rather, we surmise that aerosols emitted by continental Chilean sources were sampled during portions of the segment in Fig. 6. Vertical dashed lines indicate the subset of the 2 h segment we picked (subjectively) as being representative of onshore-moving air that was not affected, or only moderately affected, by emissions from continental Chilean sources. However, we do not expect our conditional sampling (based on HYSPLIT) and subjective picking (e.g., Fig. 6) to select aerosol properties representative of pristine marine air. Rather, we view these strategies as way to isolate aerosol properties associated with onshore-moving air that was less affected by continental sources compared to the other portions of the CCOPE data set.

Portions of three other 2 h segments were also discriminated into a period of onshore-moving air that was less affected by continental aerosols compared to an adjacent portion (or portions) of the 2 h data segment. This is shown in the Supplement. Only measurements seen plotted between the vertical dashed lines in the Supplement are analyzed in Sect. 4.3, 4.4, and 4.5.

4.3 Using N∕V ratios to parameterize cloud droplet concentration

In this section we analyze two ASD moments (Sect. 3.3). The ratio of N_UHSAS (aerosol concentration) and V_UHSAS (aerosol volume) – generically the N∕V ratio – is of interest for several reasons. First, for both operational and theoretical reasons the N∕V ratio is evaluated for particle diameters larger than ∼0.1 µm (van Dingenen et al., 2000, hereafter VD00; Hegg and Kaufman, 1998, hereafter HK98), and, importantly, the model developed to evaluate aerosol exchange between an overlying free troposphere (FT) and the marine boundary layer (MBL) successfully predicts the N∕V ratio in the MBL (VD00). Second, a value of the ratio can be derived by fitting measurements of N and V (HK98). Third, aerosol mass loading and thus an aerosol volume corresponding to an assumed particle density³ are relatively easy to evaluate. A method routinely used to evaluate aerosol mass loading involves pulling aerosol-laden air through a filter and evaluating the accumulated mass gravimetrically. Fourth, the product of an N∕V ratio and an ambient aerosol volume (aerosol mass) has been proposed as a scheme for estimating cloud droplet concentration in marine stratocumulus clouds (HK98 and VD00).

HK98 used a passive cavity aerosol spectrometer probe (PCASP) to evaluate N, V, and the N∕V ratio. Since the UHSAS counts down to a smaller diameter (0.055 µm) than the PCASP (0.12 µm), it is expected that the N∕V ratios we derive using the UHSAS will be larger than those in HK98. The main reason for this is that decreasing the lower-limit diameter increases N more than V (VD00).

As in HK98, linear least-squares regression analysis with an equation of the form $Y=a\cdot X$ was used to derive N∕V ratios. Values of N_UHSAS and V_UHSAS entered into the regressions were derived with the lower-limit diameter set at 0.055 µm (Table 3) and 0.12 µm (Table 4). The latter allows comparison to N∕V ratios in HK98. Tables 3 and 4 show the ratios and the fact that all of the Pearson correlation coefficients (r) are positive. With the exception of trajectories arriving at 12:00 UTC, 5 June, 06:00 UTC, 8 June (Table 3), and 00:00 UTC, 9 June (Table 4), all of the N∕V correlations are statistically significant at p<0.01.

Table 3Statistics for onshore trajectories (particle diameter integration in Eqs. 2, 4, and 5 is from 0.055 to 1 µm).

^a DDHHMM indicates the start and end times (day in June 2015, hour, minute) of the data segment. ^b Pearson product moment for the N_UHSAS(D=0.055 µm) on V_UHSAS(D=0.055 µm) correlation. ^c Data recording ended at DDHHMM = 080646, i.e., 14 min before the stated end time.

Download Print Version | Download XLSX

Table 4Statistics for onshore trajectories (D integration in Eqs. 2, 4, and 5 is from 0.120 to 1 µm).

^a DDHHMM indicates the start and end times (day in June 2015, hour, minute) of the data segment. ^b Pearson product moment for the N_UHSAS(D=0.120 µm) on V_UHSAS(D=0.120 µm) correlation. ^c Data recording ended at DDHHMM = 080646, i.e., 14 min before the stated end time.

Download Print Version | Download XLSX

As expected, the average N∕V ratio in the fifth column of Table 3 (417±297 µm⁻³) is larger than that in HK98 (223±76 µm⁻³). These averages are different at p=0.01. Table 4 has results based on the larger lower-limit diameter (0.12 µm). In that comparison, the Arauco N∕V ratio (159±100 µm⁻³) does not differ significantly from HK98's (i.e., p>0.01).

Application of the N∕V ratio to cloud–aerosol–precipitation modeling requires knowledge of the aerosol volume, or alternatively, knowledge of the aerosol mass loading and the aerosol particle density. The aerosol volume is then multiplied by an average N∕V ratio (e.g., the average at the bottom of the fifth column of Table 4), and their product is taken to be the modeled cloud droplet concentration (HK98 and VD00). This is straightforward, at least from the perspective of incorporating an aerosol-induced cloud feedback into a simulation, but it suffers from requiring additional information about the aerosol (aerosol volume). Because the UHSAS was unavailable for much of CCOPE (Table 1), aerosol volume is also unavailable. Another drawback is the implicit assumption that only aerosol particles larger than the lower-limit diameter (e.g., 0.12 µm in Table 4) form cloud droplets.

4.4 Using size distribution and N_CPC to parameterize CCN activation spectra

Andreae (2009) analyzed a set of aerosol concentration measurements obtained from colocated CPC and CCN instruments. Andreae's CPC measurements represent the concentration of particles no smaller than a particular diameter (∼0.01 µm; Sect. 2.2), and his CCN measurements represent the concentration of particles that activate cloud droplets at a water vapor supersaturation (SS) no larger than a particular value (Rogers and Yau, 1989; chap. 6). The latter is SS = 0.4 % in Andreae (2009).

Similar to the relationship between CCN concentration at SS = 0.4 % and CPC concentration (Andreae, 2009; his Fig. 2), we now describe how CPC and UHSAS data from CCOPE can be used to develop a function that describes CCN activation spectra. In the parameterization we develop, the independent variable is a CPC-measured aerosol concentration. While only estimates, the activation spectra we obtain represent an important step toward evaluating how CCN affected cloud and precipitation during CCOPE. We envision this assessment will be advanced when our activation spectra are used to initialize numerical models.

Our first step is to select a particle diameter, apply this as a lower-limit diameter in an integration of the UHSAS size distribution, and divide the integral by the coincident CPC-measured concentration. The result is referred to as the fractional aerosol concentration (FAC).

Figure 7a–b have graphical representations of FAC(D=0.055 µm) and FAC(D=0.120 µm).

Figure 7Two portrayals of the ASD recorded during CCOPE at 08:45:03 UTC 5 June 2015. This ASD is also plotted in Fig. 4a. Gray area in both panels represents the aerosol concentration integrated from the indicated lower-limit D to 1 µm. (a) Figure legend has the size-integrated UHSAS concentration, calculated with lower-limit D set at 0.055 µm, the CPC concentration, and the fractional aerosol concentration (FAC). (b) Figure legend has the size-integrated UHSAS concentration, calculated with lower-limit D in Eq. (2) set at 0.120 µm, the CPC concentration, and the fractional aerosol concentration (FAC).

Download

In a second step we interpret a FAC's lower-limit diameter as an upper-limit SS. We do this by applying a value for the kappa hygroscopicity parameter, which we set at κ=0.5, and by applying the kappa–Köhler formula of Petters and Kreidenweis (2007, their Eq. 6). This transformation from lower-limit D to upper-limit SS converts the FAC in Fig. 7a to FAC(SS = 0.41 %) and the FAC in Fig. 7b to FAC(SS = 0.13 %). We also evaluated how a range of the kappa parameter ($\mathrm{0.3}<\mathit{\kappa }<\mathrm{0.7}$) translates to a range of SS. Our upper-limit κ comes from airborne measurements made over the southeastern Pacific Ocean during summer (Snider et al., 2017), and our lower-limit κ is the value recommended by Andreae and Rosenfeld (2008) for simulating aerosol indirect effects over continents.

The FACs in Fig. 7a–b are two of the many available from CCOPE. One way to aggregate these is to calculate a FAC for each of the 20 onshore trajectories. For example, if we select the lower-limit diameter at D=0.055 µm, plot numerator values (Eq. 5) vs. denominator values (Eq. 5), and fit with the equation $Y=a\cdot X$, the “a” we derive is the FAC(D=0.055 µm) for a particular trajectory. FACs calculated in this way and with lower-limit D selected = 0.120 µm are presented in the seventh columns of Tables 3 and 4. Correlation coefficients presented in the eighth columns of these tables mostly exceed 0.5. By averaging over the 20 onshore trajectories, we calculated the overall averages presented at the bottom of the two tables. These overall averages are FAC(D=0.055 µm) = 0.35±0.13 (Table 3) and FAC(D=0.120 µm) = 0.13±0.07 (Table 4). This decrease in the FAC results because a larger lower-limit D (Eq. 5), implies a smaller numerator (Eq. 5) and thus a smaller FAC(D).

What we refer to as ensemble-averaged FACs were derived by selecting the numerator and denominator values represented in Eq. (5) from all 20 onshore trajectories. The selected data pairs were fitted in the manner discussed previously. In addition, upper and lower quartile values of the fitted slopes were calculated by applying the technique of Wolfe and Snider (2012; their Fig. 4d). We evaluated four ensemble-averaged FACs corresponding to four selected diameters (D=0.070, 0.095, 0.120, and 0.200 µm). The FAC at D=0.055 µm was eliminated from this analysis because Kupc et al. (2018) showed that UHSAS measurements, at D≤0.070 µm, are negatively biased. Results are presented as circles in Fig. 8 and vertical error bars represent the quartile range. Values plotted on the abscissa correspond to the four diameters, each transformed to an SS using the kappa–Köhler formula with κ=0.5, and horizontal error bars extend from most hygroscopic (κ=0.7), at the left-most limit, to least hygroscopic (κ=0.3), at the right-most limit.

Figure 8Parameterized CCN activity spectrum derived using CPC and UHSAS measurements from the 20 onshore trajectories that arrived at the Arauco site between 29 May and 28 June 2015. Pink circles and the pink fit line are for lower-limit diameters set at 0.200 and 0.120 µm. Black circles and the black fit line are for lower-limit diameters set at 0.095 and 0.070 µm. Figure legend has power-law coefficients describing the parameterization; i.e., how FAC varies with SS.

Download

In Fig. 8 we used power laws of the form FAC(SS) = C⋅SS^k (i.e., the form commonly used to parameterize CCN activation spectra; Twomey, 1959) to fit the points. The change in the slope of the fit function, seen here at SS = 0.15 %, seems consistent with analyses, demonstrating that in polluted marine cloud conditions, albeit during summertime, the exponent “k” in the Twomey power fit function is ≥1 and ≤1 at SS < 0.1 % and SS > 0.1 %, respectively (Hudson and Nobel, 2014; data from the MASE project in their Fig. 1).

Our parameterized CCN activation spectrum (Fig. 8) is relevant to cloud–aerosol–precipitation modeling for several reasons. First, some numerical models treat SS as a prognostic variable and thus require initialization with a CCN activation spectrum (e.g., Khairoutdinov and Kogan, 2000). Similarly, some models initialize with a particle-size-dependent ASD function and use Köhler theory to derive a model-initializing CCN activation spectrum (e.g., Lebo et al., 2012). As described in these two references, these models initialize with a nonspecific CCN activation spectrum. If those models were used to investigate wintertime clouds and precipitation on the central Chilean Coast, our parameterization could be applied as a CCOPE-specific initialization. Second, since we have measurements of N_CPC for the totality of CCOPE (Table 1), and we have shown how an ensemble-averaged CCN activation spectrum can be developed with N_CPC as the input parameter – i.e., as $N\left(\mathrm{SS}\right)=\mathrm{FAC}\left(\mathrm{SS}\right)\cdot N_{\mathrm{CPC}}$ – our parameterization can be used to estimate activation spectra for the complete CCOPE campaign. Third, model initiation with a specific CCN activation spectrum, as opposed to initialization with a regime-dependent droplet concentration (e.g., Thompson et al., 2004), is justified by sensitivities to cloud droplet activation reported in several publications (Cooper et al., 1997; Hudson and Yum, 1997; Snider et al., 2017).

An assumption implicit in our development is that particles were internally mixed within each of the four particle size classes. This seems justified by our use of HYSPLIT to conditionally sample (Sect. 3.1) and the fact that the sampled air masses were resident in the marine boundary layer for hours to days while subject to a variety of processes (Brownian coagulation and reactive uptake of SO₂, among others) that produce aerosols consistent with the internal mixture assumption (Fierce et al., 2017). An aspect of our measurements also supports the internal mixture assumption. Figure 7b shows that number concentration corresponding to the 0.120 to 1 µm class is dominated by particles with diameters at the lower end of that class. Hence, the contribution of freshly emitted SSA particles, generally thought to size at dry diameters larger than 0.5 µm (Clarke et al., 2003; LS04), and with a κ=1.2 (Berg et al., 1998), is typically small. A different bias would result if particles with κ values smaller than the lower-limit value (κ=0.3) contributed significantly to the size-integrated concentration in Eq. (5). Burning biomass is an important source for such low-hygroscopicity particles (Carrico et al., 2005). Our conditional sampling (Sect. 3.1), combined with our filtering of the CPC and UHSAS measurements (Sect. 3.1 and Appendix B), reduces this concern.

4.5 Regression of N_>0.5 and sea surface wind speed

As discussed in Sect. 3.2, N_>0.5 represents the concentration of particles larger than 0.5 µm. We now support our conjecture that particles grouped into the N_>0.5 subset are indeed SSA. We do this by analyzing the correlation between N_>0.5 and sea surface wind speed (U). Section 3.1 explains how we used HYSPLIT to derive U.

Values of N_>0.5, corresponding to the 18 sea surface trajectories (Sect. 3.1), are plotted against U in Fig. 9. Linear least-squares regression analysis with a model equation of the form $\ln \left(N_{>\mathrm{0.5}}\right)=\ln \left(N_o\right)+a_N\cdot U$ was used to derive the coefficients N_o and a_N (O'Dowd and Smith 1993; LS04). The fitted coefficients are N_o=0.15 cm⁻³ and a_N=0.38 s m⁻¹ and the derived function (black curve) is shown in Fig. 9. The dashed black curves represent the 95 % confidence interval (Romano, 1977; his Eq. 4.2.3.f). Also plotted (pink line) is the function derived by O'Dowd and Smith (1993) for dried SSA particles with a diameter between 0.38 and 0.84 µm. Given that the O'Dowd and Smith (1993) function (their Fig. 7a) is associated with statistical uncertainty comparable to what we estimate for our data set, we are only moderately confident that the function we derived is a consequence of wind-generated SSA. Two caveats require mentioning. First, a fraction of our data points (∼25 %) lie either above or below our confidence interval (Fig. 9). Meteorology can contribute to this variability, as when sea surface winds establish a SSA population, and the wind subsequently slacks or speeds up prior to advection onto the continent. This is expected because the atmospheric residence time of D∼0.5 µm particles, in the absence of precipitation, is several days (LS04, p. 76). Also, our unintentional sampling of particles generated over the continent is a concern. We have taken steps to eliminate those sources of contamination (Sect. 3.1 and Appendix B), but our methods are not foolproof.

Figure 9Averaged values of N_>0.5 (±1 standard deviation) vs. HYSPLIT-derived averaged Us for the 18 sea surface trajectories that arrived at the Arauco site between 29 May and 28 June 2015. The black curve is the fit of the CCOPE data; dashed curves above and below the black curves are 95 % confidence intervals (Romano, 1977; his Eq. 4.2.3.f). The pink curve is the fit reported by O'Dowd and Smith (1993) for 0.38 µm < D<0.84 µm.

Download