Hygroscopicity and composition of California CCN during summer 2010

Authors


Abstract

[1] We present an overview and analysis of cloud condensation nuclei (CCN) sampled in California by a NOAA WP-3D aircraft during the 2010 CalNex project. Four distinct geographical regions are characterized, including the Los Angeles basin, the San Joaquin and Sacramento Valleys, and the eastern Pacific Ocean west of southern California. Median size distributions in the Central Valley were unimodal (Dg ∼ 25 nm) with a larger fraction of organic species and smaller fraction of nitrate species in the Sacramento Valley aerosol than in the San Joaquin Valley aerosol. Size distributions in the Los Angeles basin and marine outflow were bimodal (geometric mean diameter, Dg ∼ 30, 90–100 nm) with similar organic fractions and some replacement of nitrate with sulfate in the marine outflow. Both fine particle and CCN concentrations were found to decrease rapidly above the planetary boundary layer (∼2 km altitude), with CCN concentrations in the boundary layer ranging from ∼102–104 cm−3 STP, while fine particle concentrations (0.004–1 μm diameters) ranged from ∼103–105 cm−3STP. The CCN-active number fraction varied between 0–100% in the Los Angeles Basin and Marine Outflow, but was substantially lower (0–40%) in the San Joaquin and Sacramento Valleys. Values of the hygroscopicity parameter,κ, inferred from the CCN measurements varied from 0.1–0.25, with the highest values in the marine outflow and the lowest values in the Sacramento Valley. The κvalues agreed well with the predictions based on size-resolved aerosol composition, but were overpredicted by almost twofold when size-averaged composition was used. CCN closure was assessed for simplified compositional and mixing state assumptions, and it was found that assuming the aerosol to be internally mixed overpredicted CCN concentrations by 30–75% for all air mass types except within the Sacramento Valley, where good closure (overprediction < 10%) was achieved by assuming insoluble organics. Assuming an externally-mixed aerosol fraction or incorporating size-resolved composition data improved closure in the other three regions, consistent with the bimodal nature of the aerosol size distribution.

1. Introduction

[2] The central and southern portions of California experience some of the most severe air quality problems in the United States, which is due to intense local emissions sources combined with topographically-influenced air flow that contains these emissions within the low-lying valleys and basins throughout the region between Los Angeles and Sacramento. Much work has been conducted in the past decades to understand the magnitude and spatiotemporal distribution of this pollution. A strong seasonal variation has been observed by multiple studies indicating that the highest particle concentrations are reached during the winter when strong temperature inversions limit vertical mixing and ventilation of the Los Angeles basin and San Joaquin valley. During the summer, a more westerly flow driven by diabatic heating in Mojave Desert and the Sierra Nevada mountains draws air inland, significantly decreasing local particle concentrations relative to those observed during winter [e.g.,Chow et al., 2006; Rinehart et al., 2006; Dillon et al., 2002; Green et al., 1992, and references therein]. The prevailing westerly winds enter the San Joaquin and Sacramento Valleys through the Carquinez Strait east of San Francisco, and move northward into the Sacramento Valley and southward into the San Joaquin Valley; infrequently, air enters the San Joaquin Valley through additional entrance points to the south. Similar to the sea breeze influence in the Los Angeles basin, these winds can induce a west-east pollution gradient that leads to higher aerosol concentrations near the upslope side of the Sierra Nevada and San Gabriel mountain ranges.

[3] It is known that increased concentrations of soluble aerosols affect cloud properties and lifetime by increasing droplet number and reducing the mean cloud droplet size [Twomey, 1977; Albrecht, 1989]. Recent work has suggested that these aerosol indirect effects on clouds can also suppress orographic precipitation on the windward side the southern Sierra Nevada mountains [Rosenfeld et al., 2008; Givati and Rosenfeld, 2004]. This implies that clouds are able to persist long enough to be transported over the crest of the mountains before either raining out or evaporating on the leeward side [Rosenfeld et al., 2008], with important implications for regional water cycling. While urban pollution plumes were found to be important sources of these enhanced aerosol concentrations, Rosenfeld et al. [2008]also report that non-urban primary and secondary aerosol sources likely play an important role.

[4] In addition to inland transport, coastal pollution sources also can be transported out over the eastern Pacific ocean and influence the structure of stratocumulus cloud decks near the coast [e.g., Hegg et al., 2009; Furutani et al., 2008; Roberts et al., 2006]. These observed effects of aerosols on clouds and precipitation motivated the need to better understand the ability of California aerosol to affect cloud droplets by contributing cloud condensation nuclei (CCN). The ability of a particle to act as a CCN depends on both its size and chemical composition, which are, in turn, dependent on aerosol microphysical processes, emissions, and atmospheric processing.

[5] Major summertime aerosol sources in the San Joaquin Valley and Los Angeles basin include secondary production of nitrate and organic material (36–38% by mass), motor vehicle emissions (13–25%), fugitive dust (16–19%), and agricultural and animal husbandry operations (4–5%) [Chen et al., 2007]. Wintertime emissions sources are more heavily affected by secondary nitrate aerosol production and residential wood burning [Chen et al., 2007]. Neuman et al. [2003]found evidence of gas-phase HNO3depletion associated with secondary gas-to-particle production of NH4NO3 aerosol in the San Joaquin Valley and Los Angeles basin during April and May, 2002, which was not observed in the Sacramento Valley or over the coastal Pacific Ocean. This highlights the importance of agricultural ammonia emissions from local animal husbandry operations as the dominant source of atmospheric bases; Sorooshian et al. [2008] and Zhang and Anastasio [2001] also observed smaller sources of organic nitrogen (e.g., amines and amino acids) in the San Joaquin Valley that could neutralize NO3. The timescales of secondary nitrate and, to a lesser extent, organic aerosol production are thought to be much less than the boundary layer mixing timescale, which result in horizontal and vertical aerosol concentration gradients in the Los Angeles basin and Central Valley [Duong et al., 2011; Neuman et al., 2003; Collins et al., 2000]. Consequently, using measurements at one or several locations to represent an entire global model grid cell (∼100 km by 100 km) can lead to substantial uncertainties [Collins et al., 2000].

[6] In this study, we seek to comprehensively characterize the compositional and size-dependence of CCN activation and hygroscopicity of California aerosol over a wide horizontal and vertical sampling area during May–June, 2010. Average regional aerosol size, composition, and hygroscopic properties are presented, which are relevant for scales typical of global models. Thus, this work will complement previous, more-focused studies of Los Angeles CCN measured near local sources [Cubison et al., 2008], of Central Valley CCN affected by the plume from a large commercial cattle operation [Sorooshian et al., 2008], of the coastal CCN gradient west of Los Angeles [Furutani et al., 2008], and of eastward CCN transport in the southern Sacramento Valley [Rosenfeld et al., 2008].

2. Observational Data Set

2.1. Study Location

[7] Measurements were performed aboard a National Oceanic and Atmospheric Administration WP-3D aircraft based in Ontario, California (34°3′10″N, 117°37′40″W). Eighteen research flights were conducted throughout the Los Angeles basin, San Joaquin and Sacramento Valleys, and coastal regions of California during the CalNex campaign from May 4th–June 20th, 2010 (Table 1). Flight trajectories are shown in Figure 1.

Table 1. Research Flights During 2010 CalNex
Research FlightDateLocal Time (PDT, UTC - 7 h)Flight Location and Description
14 May11:40–16:27Los Angeles basin
27 May10:04–16:58Transects across the Central Valley
38 May11:05–18:09Los Angeles basin
411 May10:03–17:13Rice fields near Sacramento and coastal inflow through the Carquinez Strait
512 May10:00–18:48Transects across the Central Valley
614 May10:01–16:12Los Angeles basin and upwind over the Pacific Ocean coast
716 May10:58–18:42Los Angeles basin and upwind over the Pacific Ocean coast
819 May10:27–17:11Los Angeles basin
921 May08:27–11:26Ship plume interception and upwind of the Los Angeles basin
1024 May16:07–22:00Los Angeles basin to coast and Central Valley
1130 May18:58–00:45Los Angeles basin during nighttime
1231 May21:59–03:54Los Angeles basin during nighttime
132 Jun00:58–07:08Los Angeles basin during early morning
143 Jun00:58–07:41Los Angeles basin during early morning
1514 Jun10:55–18:16Rice fields near Sacramento and inflow through the Carquinez Strait
1616 Jun11:02–17:57Transects across the Central Valley
1718 Jun11:03–18:08Transects across the Central Valley, flow through the Carquinez Strait, and westward profile of coastal emissions gradient over the Pacific Ocean
1820 Jun10:57–18:04Los Angeles basin
Figure 1.

Overview of NOAA WP-3D flights during CalNex.

2.2. Chemical Composition Measurements

[8] Non-refractory aerosol chemical composition was obtained from a compact time-of-flight aerosol mass spectrometer (C-ToF-AMS) with a pressure-controlled inlet [Bahreini et al., 2008; DeCarlo et al., 2006; Drewnick et al., 2005]. The C-ToF-AMS operates by focusing the sample aerosol stream onto a hot plate to vaporize the non-refractory aerosol components, which are subsequently ionized and detected by the time-of-flight mass spectrometer. The instrument can be operated in either “time of flight” mode or in “mass spectrum” mode, where the former involves periodically interrupting the aerosol stream with a rotating chopper to measure the size-dependent particle time-of-flight across the vacuum chamber, while the latter alternates the chopper in and out of the beam line to obtain size-averaged differential mass spectra. Both methods of operation yield mass spectra at 0.1 Hz, although, the size-resolved mass distributions were averaged over five-minute intervals to improve the signal-to-noise ratio. Mass loadings (μg cm−3) of sulfate, nitrate, ammonium, and organic aerosol constituents were then obtained from the mass spectra via the procedure of Allan et al. [2003], with relative uncertainties of ±34–38% [Bahreini et al., 2009]. The C-ToF-AMS only measures non-refractory aerosol chemical composition, so this analysis neglects the contribution of black carbon (BC), sea salt, and crustal species. Concurrent measurements of BC mass on the WP-3D show it to be much less than the non-refractory aerosol mass (<3%), while the latter species typically reside in the coarse aerosol size mode and do not contribute to fine particle number concentrations. This is supported by the good mass closure between the C-ToF-AMS total mass + BC mass and the total aerosol volume calculated from the particle size distribution to within 10–20%. Consequently, omitting these compounds from the forthcoming analysis is unlikely to have a significant effect on the observed CCN characteristics.

2.3. Particle Size Distribution Measurements

[9] The dry aerosol size distribution (0.004 to 8.3 μm diameters) was measured using an ultra-high sensitivity aerosol size spectrometer (UHSAS), a white-light optical particle counter (WLOPC), and a nucleation mode aerosol size spectrometer (NMASS). The NMASS is made up of five condensation particle counters with 0.004, 0.008, 0.015, 0.030, and 0.055μm detection diameters, and the fine particle size distributions (diameters < 1 μm) were obtained by integrating the size bins to the UHSAS distribution using a nonlinear inversion algorithm [Brock et al., 2000]. The sizing instrumentation is described by Brock et al. [2011], and the uncertainty of the fine particle concentrations is approximately ± 11% for concentrations above 100 cm−3.

2.4. CCN Measurements

[10] CCN measurements were conducted using a Droplet Measurement Technologies streamwise, thermal-gradient cloud condensation nuclei counter (CCNC) [Roberts and Nenes, 2005; Lance et al., 2006], located downstream of a 2.5-μm-cutoff-diameter impactor. The CCNC consists of a cylindrical tube with wetted walls on which a streamwise linear temperature gradient is applied. Since the diffusivity of water vapor exceeds the thermal diffusivity of air, a water vapor supersaturation is generated, which is maximum at the centerline. Particles are introduced at the column centerline, and those that activate to form droplets are counted and sized by an optical particle counter at the base of the column.

[11] The CCNC supersaturation profile mainly depends on the applied temperature gradient, flow rate, and pressure [Roberts and Nenes, 2005]. During CalNex, the instrument was operated as a CCN spectrometer using the Scanning Flow CCN Analysis (SFCA) technique of Moore and Nenes [2009]. SFCA entails dynamically scanning the instrument flow rate over time to produce a nearly-instantaneous change in supersaturation, while maintaining a constant applied temperature gradient and pressure. A flow orifice and active flow control system were employed as in work byMoore et al. [2011]to maintain a constant instrument pressure of 500 hPa, while the applied temperature gradient was kept constant at 12 K. Upscan and downscan flow ramp times of 15 seconds were used, and CCN counts were integrated over each second to obtain 1-Hz CCN concentrations.

[12] Supersaturations were calibrated in terms of the CCNC internal temperature gradient, the instantaneous flow rate, and the overall flow rate range (i.e., the minimum and maximum flow rate in each scan). The relationship between supersaturation and instantaneous flow rate was found following the procedure of Moore and Nenes [2009], where size-classified ammonium sulfate particles from a differential mobility analyzer (DMA) were introduced into the CCNC, and the DMA voltage was varied in a stepwise manner so that approximately three CCNC flow scans were obtained at each particle size. Sigmoidal activation curves of CCN versus flow rate are obtained, and the inflection point of the sigmoid is used as the critical activation flow rate,Qc, which corresponds to the critical supersaturation, sc, above which particles act as CCN. For each particle size, sc is obtained from Köhler theory [Köhler, 1936], as in work by Rose et al. [2008] and Moore et al. [2010]. The molality-dependent osmotic coefficient of ammonium sulfate used in Köhler theory is computed using the ion-interaction approach ofPitzer and Mayorga [1973] with parameters obtained from Clegg and Brimblecombe [1988]. The absolute uncertainty of the calibrated CCNC supersaturation is estimated to be ±0.04%, while the uncertainty in CCN number concentration from counting statistics and fluctuations in temperature, pressure, and flow rates during flight operation is estimated to be 7–16% for CCN concentrations above 100 cm−3 STP, which is comparable to the uncertainty of the instrument operating in constant flow mode [Moore et al., 2011]. The overall flow rate range was found to be sensitive to the large changes in the ambient air temperature (∼10°C) encountered throughout the CalNex mission and both the minimum and maximum flow rate would gradually shift toward larger values as the aircraft cabin temperature increased during flight. This temperature dependence was found to impact the calibrated supersaturations during the flow downscans, while only weakly-influencing the calibrated supersaturations during the upscans. Consequently, calibration curves for multiple flow ranges were obtained to account for these effects, and only flow upscan data are used for this analysis.

3. Results and Discussion

3.1. Regional Air Types

[13] During the eighteen research flights from May 4th–June 20th, the WP-3D aircraft sampled aerosol in four geographical regions: the Los Angeles basin, the continentally-influenced marine coastal environment west of Los Angeles (hereafter referred to as Marine Outflow), the San Joaquin Valley, and the Sacramento Valley.Figure 1 shows the aircraft flight trajectories during CalNex, as well as the geographical regions used for this analysis.

[14] The median particle size distributions and mean aerosol volume fractions (relevant for CCN activation) for each sampling region are shown in Figure 2. The distributions were obtained by computing the median aerosol concentration (and interquartile range) in each size bin for all distributions measured during CalNex. While this method produces a characteristic size distribution that is wider than any of the individual measured distributions, it does capture the relative importance of distinct aerosol modes to the regional aerosol population in terms of both their magnitude and observation frequency. Thus, we provide these distributions to inform the interpretation of aerosol and CCN characteristics, while using 30-second averaged size distributions for the calculations insections 3.2, 3.3, and 3.6.

Figure 2.

(left) Median size distributions for each geographical region. Shaded area denotes the interquartile range. (middle) Average aerosol volume fractions calculated from the C-ToF-AMS mass loadings as described insection 3.1. (right) Median aerosol ion mass loadings measured by the C-ToF-AMS. Error bars denote the interquartile range for all species. Median mass concentrations of non-refractory chloride were below the instrument detection limit of 0.03μg m−3 for all sampling regions.

[15] Distinct aerosol modes are present at 30 nm and 90 nm in the Los Angeles basin, with a similar but slightly larger bimodal distribution in the Marine Outflow from the basin (Dg = 30, 100 nm). A similar bimodal size distribution was seen by Cubison et al. [2008]for measurements in Riverside, California, during summer, 2005. They attribute small mode organics observed with an aerosol mass spectrometer to be externally-mixed with a larger, aged, internally-mixed mode of organics and inorganics [Cubison et al., 2008], consistent with measurements in other urban and non-urban settings [e.g.,Murphy et al., 2006; Zhang et al., 2005]. Median size distributions in the San Joaquin and Sacramento valleys were found to be primarily unimodal (Dg ∼ 25 nm) with a smaller secondary mode centered around Dg ∼ 80 nm.

[16] The average aerosol volume fractions shown in Figure 2were calculated from the C-ToF-AMS mass loadings assuming the aerosol to be internally mixed. Inorganic species densities were obtained from tabulated values and an organic density of 1400 kg m−3 is assumed [Lance et al., 2009]. Neutral and acidic sulfate species were differentiated using the molar ratio of ammonium ions to sulfate ions, RSO4, and mass balance as per Nenes et al. [1998]. For RSO4 > 2, sulfate is fully neutralized by the available ammonium and is present as ammonium sulfate, while for 1 < RSO4 < 2, the sulfate is present as ammonium sulfate and ammonium bisulfate. For RSO4 < 1, the sulfate is a mixture of ammonium bisulfate and sulfuric acid. Nitrate was found to constitute a significant fraction of the aerosol volume and is likely neutralized since HNO3is volatile at the ambient temperatures typically encountered during CalNex. In a limited number of cases constituting less than 10% of the sampling time in the Los Angeles basin and San Joaquin Valley, the measured ammonium mass was insufficient to fully neutralize both the measured sulfate and nitrate mass loadings. This suggests either the presence of other cations not explicitly resolved by the C-ToF-AMS (e.g., aminium salts) or the presence of externally mixed acidic sulfate. The “excess nitrate” that cannot be neutralized by ammonium does not contribute significantly to the overall aerosol compositions shown inFigure 2, but was found to comprise less than 6% of aerosol volume during some periods in the Los Angeles basin and San Joaquin Valley. Given the uncertainty of the C-ToF-AMS measurements and external mixing effects, this estimate likely represents an upper limit.Sorooshian et al. [2008] measured the aerosol composition downwind of a large bovine source in the San Joaquin Valley and found a significant contribution from amines even in the presence of ammonia. Excess nitrate mass loadings reportedly reached 0.89–1.72 μg m−3 within the source plume, but were close to zero for background conditions in the valley [Sorooshian et al., 2008]. Organic species dominate all mass types (∼49–79% by volume), with smaller contributions from ammonium sulfate and ammonium nitrate. Overall, unneutralized sulfate species constitute only 8–14% of aerosol volume on average.

3.2. CCN Activity

[17] Vertical profiles of the measured CCN concentration, NCCN, and fine particle condensation nucleus (CN, 0.004–1 μm diameters) concentration, NCN, are shown in Figure 3. Fine particle concentrations range from 500–100,000 cm−3STP, and concentrations exhibit a decreasing trend above about 1–2 km altitude. CCN concentrations follow a similar trend, but with at least ten-fold lower concentrations, typically ranging from a few tens to several thousand particles per cm3 STP. The overall trends of the vertical profiles observed in each sampling region are similar.

Figure 3.

Vertical profiles of (left) CCN concentrations at 0.25–0.65% supersaturation and (right) submicron, fine particle concentrations measured during CalNex. Data are color-coded by sampling region as inFigure 2.

[18] SFCA provides fast measurements of CCN concentrations over 0.25–0.65% supersaturation, s, during a 15-second flow scan, and these values were fit to a sigmoidal function of the form:

display math

where a0, a1, and a2 are empirical fitting constants. The fit function was then used to find the differential CCN distribution, dNCCN/dsover the range of 0.2–0.7% supersaturation for each flow upscan. This indirect procedure has the advantage of smoothing out any Poisson counting statistics uncertainty for the 1 Hz CCN concentrations, and is analogous to concentration averaging over 10-second and 30-second periods employed byMoore et al. [2011] and Asa-Awuku et al. [2011] for constant flow operation of the instrument.

[19] Figure 4shows the average CCN and particle size distributions as a function of C-ToF-AMS organic volume fraction for each sampling region. The geometric mean supersaturation of the CCN distribution varies between 0.35–0.40% for all sampling regions, while the breadth of the distribution is more variable. Most of the CCN distribution falls within 0.2–0.4% supersaturation in the Los Angeles basin and Marine Outflow regions, although broadening toward higher supersaturations occurs at the highest organic loadings. The bimodal structure seen inFigure 2 for these regions is reflected in the size distribution when the organic volume fraction, εorg is below 0.6. As εorgincreases, the size distribution transitions to a unimodal shape. This may reflect coagulation and condensational growth that blurs the distinction between a smaller mode of “fresh,” primary particles and a larger mode of well-aged background aerosol similar to that observed byCubison et al. [2008]. Although total fine particle concentrations peak at the highest εorg for the marine region, dNCCN/dsdecreases in magnitude and broadens considerably, possibly from coastal sources of organic-rich and less-CCN-active pollution aerosol.

Figure 4.

Average (left) CCN supersaturation distributions and (right) particle size distributions plotted versus the C-ToF-AMS organic volume fraction. Solid traces denote the geometric mean supersaturation and geometric mean diameter.

[20] The particle size distributions shown in Figure 4 for the San Joaquin and Sacramento Valleys are much more concentrated and unimodal. A more pronounced increase in particle size with increasing εorg is observed, which is consistent with condensational growth of the Aitken mode. The dNCCN/ds distributions in these valleys are much broader than in the Los Angeles basin, although the difference in geometric mean supersaturation between the regions is relatively small. This CCN distribution is widest at εorg∼ 0.5–0.7 in the San Joaquin Valley and 0.7–0.75 in the Sacramento Valley, suggesting a mixture of hygroscopic and less-hygroscopic aerosol modes.

[21] In summary, while there is some variation in the regional size distribution, most particles are present in the 10–100 nm size range, which constrains the range of supersaturations required for CCN activation, consistent with the well-known principle that size is more important than composition in determining CCN activity [Dusek et al., 2006; Twomey, 1977]. Composition does play an important role, however, which is reflected in the broadening of the CCN spectrum at high aerosol organic fractions. This is seen most clearly in the San Joaquin and Sacramento Valleys, particularly the latter, where the average aerosol organic fraction is greatest.

3.3. Inferring Hygroscopicity

[22] The compositional and size-dependence of CCN activation is described by Köhler theory [Köhler, 1936], and a single parameter representation of this theory has been widely-adopted in recent years [Petters and Kreidenweis, 2007]. The critical water vapor supersaturation, sc, required for a particle to act as a CCN is given by

display math

where Dp,c is the corresponding critical dry particle diameter, κ is the effective hygroscopicity parameter, R is the universal gas constant, T is the absolute temperature, σ is the surface tension of the solution droplet at the point of activation, and Mw and ρw are the molar mass and density of water, respectively. Here, the surface tension of pure water is assumed following the convention of Petters and Kreidenweis [2007] to facilitate comparison of κ values with other studies.

[23] As in work by Moore et al. [2011], equation (2) can be rearranged to state that when particles of a given κ are exposed to a constant water vapor supersaturation, those larger than the critical dry diameter, Dp,c will act as CCN. Thus, assuming the aerosol to be internally mixed, κ can be determined by finding Dp,c from integrating the the particle size distribution to match the measured CCN concentration at a given supersaturation,

display math

where NCCN is the measured CCN number concentration, Dp is the dry particle diameter, and nCN is the particle size distribution function. The derived values of Dp,c are then used in equation (2) to find κ.

[24] Figure 5shows the median, CCN-derivedκ values plotted versus Dp,cfor each supersaturation measured. Although the CCN measurements are not size-resolved, it is expected thatκ would be most characteristic of the aerosol size range near Dp,c, since that is where CCN concentrations would be most sensitive to changes in κ. The error bars in Figure 5 denote the interquartile range of observed values. Also shown for comparison are κvalues calculated from the size-resolved and bulk (i.e., size-averaged) C-ToF-AMS compositions as

display math

where εi and κiare the volume fraction and pure-component hygroscopicity of speciesi, respectively. As discussed in section 3.1, this calculated κneglects the contribution of refractory aerosol species not measured by the C-ToF-AMS, but whose omission is not expected to significantly affect the calculatedκ. Pure-component inorganicκ values are computed as κi = (Mw/ρw)(ρi/Mi)εi, where ρi, Mi, and νi are the density, molar mass, and van't Hoff factor of species i, respectively. An organic κ, or κorg, of either 0 or 0.11 is assumed, where the former corresponds to an insoluble organic species and the latter corresponds to a soluble organic species with, e.g., Morg of 0.200 kg mol−1 and ρorg of 1400 kg m−3. The C-ToF-AMS size-resolved compositions yield much better agreement with the CCN-derivedκthan is achieved with the bulk compositions. Assuming the organic species to be insoluble or slightly soluble has a small effect on the size-resolvedκpredictions, but the CCN-derivedκseems to agree best with the predictions based on insoluble organics. Since this calculation assumes the aerosol to be internally-mixed, this result may also be caused by an externally-mixed aerosol with both fresh (non-hygroscopic) and aged (more hygroscopic) organic species. Aerosol mixing state influences will be considered in the CCN closure study insection 3.6. Hygroscopicities derived from bulk C-ToF-AMS chemistry significantly overpredict those measured by the CCNC by almost twofold, which is consistent withCubison et al. [2008], who found size-dependent aerosol compositions were necessary to accurately reproduce the CCN concentrations observed in Riverside, CA.

Figure 5.

Aerosol hygroscopicity, κ, inferred from the measured CCN concentrations and aerosol size distributions, plotted against the critical activation diameter, Dp,c for each supersaturation (0.25–0.65%)(squares). Shown for comparison are κvalues calculated from size-resolved (diamonds) and size-averaged (circles) C-ToF-AMS measurements plotted versus particle diameter,Dp. Markers denote the median values for each air mass type, while error bars denote the interquartile range. The solid line shows the median particle size distribution from Figure 2.

3.4. Organic Oxygenation

[25] Figure 5shows that the CCN-derivedκis more consistent with size-resolved C-ToF-AMS predictions assuming insoluble organics, although mixing state effects would be expected to also play a role. Past studies have found organic aerosol hygrosocopicity to vary with the degree of oxygenation [e.g.,Jimenez et al., 2009; Chang et al., 2010; Duplissy et al., 2011; Lambe et al., 2011], with more-oxidized organics expected to be less-volatile and more hygroscopic. The C-ToF-AMSm/z 43 and 44 peaks can be used to characterize the organic oxygenation since m/z 44 is mostly the CO2+ fragment of highly oxygenated organics (e.g., dicarboxylic acids and esters) and m/z 43 is mostly C3H7+ and C2H3O+ fragments. The ratio of each peak to the total organic aerosol mass (f44 and f43, respectively) is then a proxy for the total oxygenation of the organic aerosol with higher f44 values correlated with higher O:C [Aiken et al., 2008; Zhang et al., 2005] and hygroscopicity [Jimenez et al., 2009].

[26] Figure 6 shows the 95% simultaneous confidence ellipses for f43 and f44 for each sampling region, where the center of each ellipse denotes the mean f43 and f44 values. The data were filtered to include only organic aerosol loadings greater than 1 μg m−3to improve signal-to-noise; while reducing the overall scatter for some regions, this filtering process had a negligible effect on the statistical means. Also shown is the triangular bounding region reported byNg et al. [2010]that describes the range of observations for the positive-matrix-factorization-resolved oxygenated organic aerosol (OOA) factor of ambient and chamber data.Ng et al. [2010] define fX as the ratio of m/z X to only the OOA aerosol mass, while this study uses the total organic mass. These different definitions are unlikely to bias the comparison between the range of f44 values, but would be expected to shift the f43 values in this study to higher values than seen by Ng et al. [2010] since m/z43 incorporates both oxygenated and non-oxygenated fragments.

Figure 6.

Oxidation state of the aerosol expressed as f44 versus f43, following Ng et al. [2010]. Points denote the 30-second-averaged observations for each sampling region, and the ellipses denote the 95% confidence regions. Only the subset of the C-ToF-AMS data where the organic mass loading exceeded 1μg m−3are included to improve signal-to-noise, and the number of points before and after filtering are shown in the legend. Dashed lines are the parameterized triangular bounding region reported byNg et al. [2010] for SOA chamber oxidation and ambient measurements. The ordinate κorg scale was calculated from the parameterizations of Aiken et al. [2008] and Lambe et al. [2011] relating f44 to O:C and κorg.

[27] It can be seen from Figure 6 that the organic oxygenation of aerosol in the Los Angeles basin and San Joaquin and Sacramento Valleys is less variable than that observed in the marine outflow. Typical f44 values were ∼0.1–0.2, which corresponds to an approximate O:C ratio of ∼0.45–0.85 [Aiken et al., 2008] and an organic κ of ∼0.1–0.2 [Lambe et al., 2011]. This CCN-derivedκorg is at the lower end of the range of 0.1–0.3 reported by Hersey et al. [2011]for ground-based measurements in Pasadena, CA, during 2009.Hersey et al. [2011] also report organic aerosol O:C ∼0.44–0.55, consistent with the lower end of observations during CalNex. The range of organic κ derived in this study from the f44 parameterizations of Aiken et al. [2008] and Lambe et al. [2011]agrees very well with the CCN-derivedκ in Figure 5, reflecting the dominance of organic species at small aerosol sizes. An open question is whether the O:C ratio derived from size-averaged C-ToF-AMS measurements is truly representative of the smaller Aitken-mode particles that determine CCN activity. These results suggest that approximating the CCN-sensitive small aerosol mode with size-dependent composition and size-averaged organic oxygenation properties gives a reasonable prediction of the observed CCN hygroscopicity.

3.5. Sensitivity of CCN to Composition Effects

[28] While quantifying the aerosol hygroscopicity is important for parameterizing CCN activation and growth, it is also important to assess the overall sensitivity of CCN to κ. To do this, we examine the theoretical sensitivity of the CCN activated ratio, Ra, which is defined as the CCN concentration normalized by the total fine particle concentration. Following Wang et al. [2008] and Moore et al. [2011], we assume that the aerosol size distribution for each region (shown in Figure 2) can be represented well by the sum of two lognormal aerosol modes. The sensitivity of Ra is then given as

display math

where nCN is the size distribution function, evaluated at Dp,c, as

display math

and Dg,i and σg,i are the geometric mean diameter and geometric standard deviation of each mode, i. Since Dp,c is related to κ and sc in equation (2), we can express ∂ Ra/∂ lnκ across the range of possible supersaturations. This is shown in Figure 7 for each sampling region, using bimodal fits to the median size distributions shown in Figure 2.

Figure 7.

Sensitivity of the activated ratio, Ra to κ as a function of supersaturation. Curves for each sampling region were computed assuming an internal mixture and constant values of κ, while using a bimodal fit to the median size distributions shown in Figure 2.

[29] As discussed by Moore et al. [2011], the shape of ∂ Ra/∂ lnκ implies that Ra is most sensitive to composition changes when Dp,c is near the maximum of the size distribution. As discussed in the preceding section, California CCN exhibit κ ∼ 0.1–0.2, and the sensitivity curves in Figure 7 for these values suggest that aerosol in the Los Angeles basin and Marine Outflow are at peak sensitivity to compositional effects between 0.25–0.4% supersaturation, while aerosol in the Sacramento and San Joaquin Valleys are more sensitive to κ above 0.4–1% supersaturation.

[30] These estimates are based only on the measured, median size distributions for each sampling region. Yet, the low end of supersaturations where Ra becomes sensitive to κagrees very well with the range over which SFCA-measureddNCCN/ds values are highest (Figure 4). This is particularly true for the Los Angeles Basin and Marine Outflow regions, while poorer agreement is seen in the San Joaquin and Sacramento Valleys. This implies that California convective and stratiform clouds with supersaturations on the order of 0.2–0.4% are particularly sensitive to aerosol chemical composition effects. This analysis also implies that much of the average CCN supersaturation distribution can be captured assuming a constant aerosol size distribution and a constant value of κ ∼ 0.1–0.2.

3.6. CCN Closure

[31] In addition to quantifying the measured size and compositional impacts on CCN activity in terms of κ, we also seek to quantify the uncertainty associated with using common simplifying assumptions typical of those in large scale models. Termed “CCN closure”, this type of error analysis has been performed for a wide range of urban and rural sites [e.g., Asa-Awuku et al., 2011; Moore et al., 2011; Rose et al., 2010; Ervens et al., 2010; Wang et al., 2010; Bougiatioti et al., 2009; Lance et al., 2009; Cubison et al., 2008; Broekhuizen et al., 2006; Rissler et al., 2004; VanReken et al., 2003].

[32] The CCN concentrations are computed following work by Moore et al. [2011], where the C-ToF-AMS compositional data are first used to find the volume fractions of organics, ammonium nitrate, ammonium sulfate, ammonium bisulfate, and sulfuric acid. These volume fractions are then used to find the aerosolκ (equation (4)) and the critical activation diameter, Dp,c, above which all particles act as CCN (equation (2)).

[33] In applying these equations, it is necessary to make assumptions regarding the aerosol mixing state (e.g., internal versus external), organic hygroscopicity (e.g., κorg = 0 or κorg= 0.11), and whether the aerosol composition varies with particle size (e.g., size-dependent or size-invariant). In this study, we examine seven scenarios as follows:

[34] 1. Ammonium Sulfate: All particles are composed of ammonium sulfate (κ = 0.6).

[35] 2. Internal Mixture, Soluble Organics: All particles have the same composition as determined by the size-averaged, C-ToF-AMS-derived volume fractions. Organics are soluble withκ = 0.11.

[36] 3. Internal Mixture, Insoluble Organics: All particles have the same composition as determined by the size-averaged, C-ToF-AMS-derived volume fractions. Organics are insoluble withκ = 0.

[37] 4. External Mixture, Soluble Organics: Particles are composed of pure components (e.g., organic particles, ammonium sulfate particles, etc.), and the number of each type is determined by the size-averaged, C-ToF-AMS-derived volume fractions. Organics are soluble withκ = 0.11.

[38] 5. External Mixture, Insoluble Organics: Particles are composed of pure components (e.g., organic particles, ammonium sulfate particles, etc.), and the number of each type is determined by the size-averaged, C-ToF-AMS-derived volume fractions. Organics are insoluble withκ = 0.

[39] 6. Internal Mixture, Size-Dependent Composition, Soluble Organics: Particles in each size distribution bin have the same composition as determined by the size-resolved, C-ToF-AMS-derived volume fractions, but the particle compositions in different size bins may not be the same. Organics are soluble withκ = 0.11.

[40] 7. Internal Mixture, Size-Dependent Composition, Insoluble Organics: Particles in each size distribution bin have the same composition as determined by the size-resolved, C-ToF-AMS-derived volume fractions, but the particle compositions in different size bins may not be the same. Organics are insoluble withκ = 0.

[41] Closure was assessed for each scenario in terms of a CCN prediction error ratio, Φ = NCCN,predicted/NCCN. CCN predictions tended toward overprediction, and the mean percent overprediction (Φ − 1) × 100% for each sampling region and instrument supersaturation is shown in Table 2. The scenarios are arranged in columns with increasing complexity from left to right. Assuming the aerosol to be pure ammonium sulfate substantially overpredicts CCN concentrations in all sampling regions, while a modest improvement is gained by incorporating C-ToF-AMS compositions, assuming an internally-mixed aerosol population. Good agreement is found for internally-mixed aerosol with insoluble organics in the Sacramento Valley (Φ ∼ 0.92–1.01); although, given the large measured organic volume fractions and expected sources of secondary organic aerosol in this region, assuming the organics to be insoluble seems unlikely.

Table 2. Percent Overprediction (Φ − 1) × 100% in CCN Number Concentration From Different Organic Solubility and Mixing State Assumptionsa
Supersaturation (%)NAmmonium SulfateInternal MixtureExternal MixtureSize-Dependent Internal Mixture
Soluble OrganicsInsoluble OrganicsSoluble OrganicsInsoluble OrganicsSoluble OrganicsInsoluble Organics
  • a

    Data from the CCNC, C-ToF-AMS, and size distributions were averaged over 30-second periods, and N reflects the number of data points used to calculate each mean and standard deviation.

Los Angeles Basin
0.33 ± 0.041940125 ± 20473 ± 12954 ± 11751 ± 1100 ± 8537 ± 8618 ± 85
0.38 ± 0.04426893 ± 11054 ± 6040 ± 5535 ± 53−13 ± 4427 ± 406 ± 43
0.43 ± 0.04430683 ± 8849 ± 4236 ± 3733 ± 37−18 ± 3526 ± 336 ± 35
0.48 ± 0.04432383 ± 8251 ± 4639 ± 3836 ± 37−18 ± 3530 ± 3410 ± 35
0.53 ± 0.04434485 ± 7754 ± 4542 ± 4140 ± 40−17 ± 3635 ± 3516 ± 36
0.58 ± 0.04421884 ± 7156 ± 4145 ± 4142 ± 36−16 ± 3739 ± 3620 ± 37
0.63 ± 0.04409876 ± 6349 ± 5739 ± 3737 ± 49−20 ± 3435 ± 3418 ± 35
 
San Joaquin Valley
0.33 ± 0.04591152 ± 14575 ± 10745 ± 8857 ± 928 ± 5948 ± 8628 ± 75
0.38 ± 0.041283143 ± 27875 ± 23047 ± 18356 ± 1914 ± 9142 ± 7023 ± 68
0.43 ± 0.041303139 ± 31072 ± 26847 ± 22057 ± 2302 ± 9742 ± 7023 ± 65
0.48 ± 0.041314138 ± 11570 ± 9046 ± 7956 ± 801 ± 4943 ± 6623 ± 62
0.53 ± 0.041320143 ± 9874 ± 7049 ± 6360 ± 643 ± 4648 ± 6328 ± 60
0.58 ± 0.041163140 ± 8668 ± 4743 ± 4256 ± 481 ± 4742 ± 5924 ± 60
0.63 ± 0.041037136 ± 8463 ± 4437 ± 3851 ± 48−1 ± 4937 ± 5819 ± 59
 
Sacramento Valley
0.33 ± 0.04132389 ± 53736 ± 60−2 ± 4421 ± 53−56 ± 3310 ± 59−14 ± 77
0.38 ± 0.04319195 ± 21524 ± 31−6 ± 2911 ± 26−59 ± 257 ± 37−19 ± 51
0.43 ± 0.04320140 ± 10119 ± 25−8 ± 2511 ± 22−60 ± 214 ± 31−20 ± 43
0.48 ± 0.04317124 ± 7822 ± 24−4 ± 2314 ± 21−59 ± 207 ± 29−15 ± 39
0.53 ± 0.04320120 ± 7225 ± 24−1 ± 2318 ± 21−58 ± 209 ± 28−10 ± 35
0.58 ± 0.04318116 ± 6328 ± 231 ± 2221 ± 21−57 ± 2113 ± 29−6 ± 38
0.63 ± 0.04321105 ± 5524 ± 22−2 ± 2117 ± 20−58 ± 208 ± 27−13 ± 45
 
Marine Outflow Near LA Basin
0.33 ± 0.0465781 ± 9750 ± 8037 ± 7631 ± 71−8 ± 5720 ± 621 ± 61
0.38 ± 0.04145561 ± 8833 ± 4322 ± 3917 ± 38−21 ± 339 ± 42−10 ± 41
0.43 ± 0.04150556 ± 10429 ± 3220 ± 2817 ± 30−24 ± 298 ± 26−11 ± 31
0.48 ± 0.04146655 ± 9029 ± 3120 ± 2618 ± 30−−25 ± 299 ± 26−8 ± 31
0.53 ± 0.04146458 ± 10531 ± 3223 ± 2621 ± 30−24 ± 2913 ± 25−4 ± 29
0.58 ± 0.04146358 ± 7734 ± 3225 ± 2624 ± 30−23 ± 3016 ± 250 ± 29
0.63 ± 0.04145753 ± 6629 ± 3021 ± 2421 ± 28−26 ± 2814 ± 24−2 ± 27

[42] Treating the aerosol as externally mixed overpredicts CCN concentrations if organics are assumed to be soluble and underpredicts CCN concentrations if organics are assumed to be insoluble. The San Joaquin Valley is an exception with good closure (Φ ∼ 0.99–1.08) obtained for the insoluble organics case. Finally, incorporating size-dependent compositions improves closure to within ±10–25% for most sampling regions, which is similar to predictions using size-averaged composition data (Φ ∼ 0.8–1.2), despite the increased complexity.

[43] Figure 8shows the variation of prediction error ratios with measured CCN concentration, and it can be seen that the internally-mixed scenario assuming soluble organics consistently overpredicts CCN over the range of concentrations (Φ ∼ 1.1-2.0), while the externally-mixed, insoluble organics scenario gives a lower overall Φ but with more scatter (∼0.5–1.5). For both scenarios, the median prediction error ratio increases with decreasing CCN concentration beyond what can be explained by decreased CCN counting statistics. Some of this uncertainty may be associated with increased C-ToF-AMS composition uncertainties at low particle concentrations. The concentration dependence also appears to be slightly more pronounced for the external mixture, insoluble organics scenario than for the internally-mixed, soluble organics case. The greatest underpredictions are seen for the Sacramento Valley, where CCN concentrations were highest whenεorg was high (Figure 4). Consequently, the significant underprediction observed at high concentrations may occur from assuming insoluble organics.

Figure 8.

CCN prediction error for 0.30–0.65% supersaturation versus measured CCN concentrations. Predictions were computed assuming that the aerosols are (left) internally mixed with κorg = 0.11 and (right) externally mixed with κorg= 0. Markers denote the median error for each sampling region, while the bars denote the interquartile range. The dotted line at Φ = 1 denotes perfect agreement and the solid bounding curves indicate the CCN measurement uncertainty. The inset plots show the frequency distribution of Φ for all 30-second-averaged data points.

[44] This analysis shows that the assumed aerosol mixing state and organic solubility are important for predicting CCN in California, and that the aerosol are likely to be at least partially externally mixed with both soluble and insoluble organics; however, it is not possible to deconvolute these effects, lacking size-resolved CCN measurements.Asa-Awuku et al. [2011] and Wang et al. [2010] were able to achieve closure to within similar uncertainties (∼20%) for urban aerosol in the vicinity of Houston, TX, and Mexico City, Mexico, respectively. The latter study notes that agreement was predicated on most aerosol having κ > 0.1 [Wang et al., 2010]. 2008Cubison et al.[2008] performed CCN closure for aerosol sampled in Riverside, California, and report substantially higher prediction uncertainties (Φ ∼ 4–6) for similar size-averaged scenarios as employed here [Ervens et al., 2010], which were attributed to the influence of fresh emissions of non-CCN-active elemental carbon and small particles [Cubison et al., 2008]. Ship-based measurements in the ship channel near Houston, TX, also showed large CCN overpredictions (Φ∼1.7–2.4) in a recent closure study, which again were explained by close proximity to local emissions sources [Quinn et al., 2008; Ervens et al., 2010]. Thus, while these results are broadly representative of the regional aerosol near Los Angeles and the Central Valley of California, they do not seem to capture the increased uncertainty associated with localized fresh emissions sources that may be important for assessing air quality impacts. Given the relatively low resolution of large scale models, however, the more regional nature of these measurements may be more appropriate for future assessments of climate prediction uncertainties.

4. Summary and Conclusions

[45] Measurements of aerosol size, chemical composition, and CCN-derived hygroscopicity obtained during the CalNex project in May-June, 2010, are presented and analyzed. Assuming the aerosol to be internally-mixed was found to significantly overpredict CCN concentrations by 30–75% for all sampling regions except the Sacramento Valley, where good closure (overprediction < 10%) was achieved assuming insoluble organics. Assuming the aerosol to be externally-mixed with insoluble organics underpredicted CCN concentrations, on average. This suggests that California aerosol is likely to be only partially externally mixed, which is consistent with the observed bimodal size distributions and with the coexistence of both fresh and aged aerosol.

[46] We also quantify the compositional dependence of CCN activity in terms of the hygroscopicity parameter, κ, which was found to vary between 0.1–0.2 with very little supersaturation dependence. CCN-derivedκand those calculated from size-resolved C-ToF-AMS compositions were found to agree very well, although using size-averaged C-ToF-AMS compositions overpredictedκby almost twofold. This suggests a size-dependence toκ which is most apparent when comparing the Aitken and accumulation mode κ. Calculations based on the median size distributions for each sampling region suggest that CCN concentrations are highly sensitive to compositional effects for supersaturations above 0.3–0.4%, which agrees remarkably well with the measured dNCCN/ds distributions, which were also centered in this range. This suggests that using regional aerosol properties is sufficient for capturing the overall characteristics of CCN; although, this likely does not account for small scale features such as fresh emissions plumes.

Acknowledgments

[47] We acknowledge support from NOAA, NASA, and an NSF CAREER award. R. Moore acknowledges support from a NASA Earth and Space Science Graduate Research Fellowship. The CalNex project was supported by NOAA's Climate Change and Air Quality Programs.