Measurements and samples were collected from 14 January to 9 March 2015 at Bodega Marine Laboratory (BML; 38∘19′ N, 123∘4′ W) in Bodega Bay, California. Measurements were collected continuously unless otherwise noted. BML is located south-southwest of the northern California coastal mountain ranges and north of Point Reyes National Seashore (Fig. 1). More detailed information regarding instrumentation deployed to BML follows below.

The sampling site at BML included two instrumented trailers located ∼ 100 m east-northeast of the seashore and ∼ 30 m north of the northernmost BML permanent building. The beta attenuation monitor (BAM), for measuring particulate matter mass for particles smaller than 2.5 µm (PM2.5), and Interagency Monitoring of Protected Visual Environments (IMPROVE; Eldred et al., 1997) filters for collecting PM2.5 and particles smaller than 10 µm (PM10) were placed at an ancillary site ∼ 15 m southwest of the trailers. The IMPROVE samples are not used in the analysis that follows. Aerosol composition and ice-nucleating particle concentrations were measured in the trailer owned by the California Air Resources Board (CARB) and operated by the University of California, San Diego (UCSD). Aerosol sizing, gas-phase tracer concentrations, black carbon mass, and cloud condensation nuclei concentrations were measured in the trailer owned by the National Park Service (NPS) and operated by Colorado State University (CSU). Ambient aerosol particles were collected nearby using filters for subsequent laboratory measurement of drop freezing temperature. A more extensive description of these instruments and the data processing, quality control, and archiving is given by Petters et al. (2017). A meteorology station operated by the National Oceanic and Atmospheric Administration's Earth System Research Laboratory (NOAA/ESRL) was located ∼ 100 m north of the two trailers. The NOAA/ESRL 449 MHz wind profiling radar, radio acoustic sounding system (RASS), and 10 m surface meteorology tower were operated during the two study periods and are used for the analysis presented in later sections. Descriptions of these instruments and the NOAA/ESRL Bodega Bay meteorology station (BBY) can be found in White et al. (2013).

Size-resolved aerosol composition at Bodega Bay was measured with an aerosol time-of-flight mass spectrometer (ATOFMS) and an ultrafine aerosol time-of-flight mass spectrometer (UF-ATOFMS). The UF-ATOFMS used a diluting stage with an approximate dilution of 5× to increase aerosol ionization efficiency during periodic high particle concentrations. The use of these instruments in tandem allows the direct measurement of single-particle composition for particles in the aerodynamic diameter range 0.15–3.0 µm. The design and operating principles of these instruments have been described elsewhere (Gard et al., 1997; Su et al., 2004); thus we provide only a short overview here. Particles were dried prior to introduction to the instrument with silica-diffusion dryers in order to improve the ionization efficiency and thus improve the acquired spectra quality. Particles enter the vacuum system through either a converging nozzle or an aerodynamic lens, wherein they are accelerated to their terminal velocities. After reaching this terminal velocity, they enter the sizing region, where they travel through two continuous wave laser beams (diode-pumped, Nd : YAG at 532 nm) separated by a vertical distance of 6 cm and oriented orthogonal to each other. Because the distance is known, particle velocity can be determined by measuring the time difference between the two scattering signals. These velocities can be converted to vacuum aerodynamic diameter (Dva) through a calibration curve generated with polystyrene latex spheres (PSLs). The velocity is also used to calculate the time when a particle will arrive in the ion source region. Upon arrival in the source region, a 266 nm Nd : YAG laser is triggered to fire upon the particle, desorb it, and generate positive and negative ions whose mass spectrum are measured with a dual-polarity time-of-flight mass spectrometer.

Aerosol size distributions at BML were measured using a scanning mobility particle sizer (SMPS, TSI Inc. Model 3936) and an aerodynamic particle sizer (APS, TSI Inc. Model 3321). The SMPS was operated with a pump flow of 0.3 L min-1 and a sheath flow of 3.0 L min-1 so that the dynamic size range varied from 13.6 to 736.5 nm. The APS operated with a sample flow rate of 1.0 L min-1 and a sheath flow of 4.0 L min-1 and measured particle sizes from 0.6 to 20 µm. The APS was externally calibrated using PSLs. Particles were not dried prior to sizing, but the relative humidity (RH) of the sample line was monitored with a RH sensor (Vaisala, HMP110). The RH in the sample line varied from 7.5 to 54.7 % during measurement periods with a mean value of 35.5 %. Nearly 97 % of the sampling periods occurred while sample line RH was below 47 % (0.4 % of local (LOCAL), 4.7 % of control (CTL), and 0 % of PGF; refer to Sect. 3.1): the efflorescence humidity of sodium chloride particles (Gupta et al., 2015). Because the RH was below the efflorescence point of sodium chloride (chosen as a proxy for sea spray) for > 95 % of the sampling periods, the humidity likely had little bearing on the composite size distributions used in this analysis. Assuming spherical particles, the measured mobility diameter (Dm) is equivalent to physical diameter (Dp). APS measurements were adjusted from aerodynamic diameter (Da) to Dp assuming spherical particles and an effective density of 1.8 g cm-3. Both APS and SMPS size distributions were combined to 10 min mean distributions from their operational 1 and 5 min scan frequency.

PM2.5 mass was determined using a BAM (Met One Instruments Inc., Model BAM 1020). The mass was recorded hourly. BC mass concentration and attenuation were determined with a seven-channel aethalometer (Magee Scientific Corp., Model AE16-ER-P3-F0), operating in AE-30 mode.

Size-resolved (also referred to as “diameter scan”) cloud condensation nuclei concentrations were measured using a streamwise cloud condensation nuclei counter (Droplet Measurement Technologies Inc., CCN-100) coupled with an SMPS. The SMPS (TSI 3080 long column) was operated at a sheath-to-sample flow rate of 5 to 1.3 L min-1. Raw counts were recorded in Labview and inverted as described previously (Nguyen et al., 2014; Petters et al., 2009). The inversions account for temperature and pressure changes inside the DMA, contribution of multiply charged particles, and losses through the inlet system. The CCN was operated at a sample flow rate of 0.3 L min-1 and sheath-to-sample flow ratio of 10 : 1. Supersaturation is calibrated using atomized, dried ammonium sulfate aerosol (Christensen and Petters, 2012). The control software cycles through an automated 12-point sequence varying supersaturation between 0.06 and 0.67 %. Activation diameters are obtained for normalized activation curves (Petters et al., 2009), and the apparent hygroscopicity parameter at standard state, κ, is calculated from the supersaturation and activation diameter (Christensen and Petters, 2012; Petters and Kreidenweis, 2007).

Concentrations of gas-phase pollutants were determined using a suite of gas-phase instruments, collocated with the aethalometer, CCN counter, and the sizing instruments. A NO–NO2–NOx analyzer (TEI Inc., Model 42C), ozone analyzer (TEI Inc., Model 49C), sulfur dioxide analyzer (TEI Inc. Model 43C), and carbon monoxide monitor (Horiba Inc., APMA-370) were all utilized in this study. Gas-phase measurements were recorded every second and converted to 10 min mean time resolution.

Level 2 MODIS cloud products (Platnick et al., 2003) are used to estimate the range of marine stratocumulus cloud optical depth offshore from BML during PGF episodes with clear sky above low clouds, and to verify that the clouds in nearshore MODIS scenes are low-level cumulus or stratocumulus clouds.

Composite aerosol size distributions, trace gas and aerosol type concentrations, indicators of secondary chemical aging, and CCN activation spectra corresponding to PGF periods and control periods are derived as a primary tool for addressing the hypotheses posed in this study. Herein, we define a control period (CTL) to be any hourly period which does not fit the definition for flow through the Petaluma Gap arriving at BML found in N06 (hereafter mPGF) and does not occur during short-lived episodes of concentrated local anthropogenic pollution. In this study, mPGF periods that also meet a minimum threshold for concentrated non-local anthropogenic pollution will be called PGF. Observed causes of local anthropogenic pollution included nearby brush fires, vehicle activity at BML, and “sea breeze resampling”. During the third of those, high concentrations of anthropogenic pollution either from a local source or from further inland that was previously transported offshore returned to the measurement site via the afternoon sea breeze. Since the polluted air mass may have up to 18 h of modification by the nearby BML marine environment just before sea breeze resampling episodes, these were classified as local anthropogenic pollution and were removed from the PGF and CTL composites.

We followed the methodology of N06 in identifying Petaluma Gap flow using the BBY 449 MHz vertically profiling radar and 10 m anemometer (see Sect. 4a in N06). Briefly, this methodology includes identifying continuous periods at least 6 h in length during which wind speed and direction criteria are met both at the surface (10 m anemometer at BML) and in the lowest retrieved layer (approximately 100 to 350 m m.s.l.) of the BML 449 MHz radar. When the conditions from the N06 methodology were met, we declared the period mPGF. It is important to note that, while 449 MHz wind profiles are collected hourly, all other data from the study are collected more frequently; therefore we classified local conditions in hourly intervals.

To choose local conditions based on an indicator of anthropogenic pollution, we examined CalWater 2015 observations of CO, NOx, SMPS number concentration integrated from 13.6 to 736.5 nm (CN), and PM2.5 (collectively, peripherals). During mPGF, the interquartile range of CO and NOx was higher than the interquartile range of the same measurements during periods that did not fit mPGF (not shown). In addition, for indicators of fine particulate concentration (CN, PM2.5), the median value during PGF is higher than the upper quartile value during all periods. For all anthropogenic pollutant indicators, the maximum observation and much of the upper quartile range are higher than any observation taken during mPGF, and occurred during local anthropogenic pollution periods. An example is shown in Fig. 2, which contrasts lower tropospheric horizontal wind and virtual potential temperature (θv), selected peripheral measurements, and EC and SS particle (Table 2) subtypes (see Sect. 3.3 for particle typing, later this section) during a period when mPGF was not observed nor were anthropogenic pollutants high (Fig. 2a), a period when it is suspected local anthropogenic pollutants were sampled (Fig. 2b), and a mPGF period (Fig. 2c). Note that, during the local pollutant episode, CN and CO were elevated to the same levels as during the mPGF period for a few hours. The onset of the elevated-pollutant period occurs near the maximum in onshore sea breeze (NNW 200 m wind near 03:00 UTC on 12 February). Pollutant concentrations decrease again a few hours after local sunset, when the offshore land breeze become established. The period 03:00–11:00 UTC on 12 February 2015 is an example of a sea breeze resampling period.

In order to exclude local or sea breeze resampled anthropogenic pollutants from CTL periods, we imposed an additional requirement based upon CO concentration – hourly mean CO concentration must be above the CalWater 2015 mean plus two standard deviations (138.1 ppbv). Along with mPGF, this requirement forms the basis of a decision table (Table 1) that allows the separation of CalWater 2015 measurements into four composites. We choose CO concentration as our additional discriminator because its interquartile range during mPGF is entirely above the interquartile range from all other periods, because its overall variability is the lowest among peripheral measurements, and because elevated near-surface CO concentration was observed by aircraft during a PGF event reported in N06. Table 1 allows the compositing of observational period by PGF (mPGF and elevated CO conditions met), CTL (neither mPGF nor elevated CO conditions met), LOCAL (mPGF not met, elevated CO met), and diffuse (mPGF met, elevated CO not met). In this light, Fig. 2a–c can be seen as examples of CTL, LOCAL, and PGF periods, respectively.

The attenuation recorded by the aethalometer was used to derive the aerosol absorption coefficient (σATN) at the instrument's native 5 min resolution following the method described in Collaud Coen et al. (2010) (see their Eq. 2). This value was not corrected and thus not reported directly, as techniques for correcting aethalometer measurements require a coincident multi-channel measurement of aerosol scattering in order to determine backscattering by the filtered particles, and this additional measurement was not taken during CalWater 2015. It is noted that values of σATN generally fall in the range reported by previous studies in continental regions (e.g., Table 1 in Chung et al., 2012). The hourly mean aerosol absorption coefficient in the channels 470, 520, 590, and 660 nm were used to derive the Ångström absorption exponent (AAE), using the relation σATN(λ)=C0λ-AAE. AAE is calculated by regression to the uncorrected σATN(λ). Note that Weingartner et al. (2003) estimated that errors in σATN are only a very weak function of wavelength in the channels used; thus it is expected that instrument errors do not contribute significantly to the estimate of AAE.

ATOFMS and UF-ATOFMS can provide information on size and chemical composition (via mass spectra) for an individual particle. Generally, positive spectra reveal particle source, while negative spectra provides information on the atmospheric processing a particle has undergone (Guazzotti et al., 2001; Sullivan et al., 2007; Prather et al., 2008). ATOFMS, but not UF-ATOFMS, spectra were filtered for periodic radio frequency interference caused by a sub-optimally operating instrument component. A total of 115 416 particles were scattered and hit during PGF events, and 1 835 387 during the control time periods (see Sect. 3.1 for definition of PGF and control periods).

Single-particle spectra and size data were loaded into Matlab (The MathWorks, Inc.) and analyzed via the software toolkit YAADA (http://www.yaada.org/). Particles were divided into clusters based on their mass spectral features via an adaptive neural network (ART-2a, vigilance factor of 0.8, learning rate of 0.05, and 20 iterations with a regroup vigilance factor of 0.85) (Rebotier and Prather, 2007; Song et al., 1999). Greater than 95 % of ART-2a analyzed particles were grouped into 11 types based upon their characteristic mass spectra and size distributions. Similar to previous field studies using ATOFMS (Sullivan et al., 2007; Pratt and Prather, 2009; Cahill et al., 2012; Qin et al., 2012), particle types were assigned by a human operator based upon similarity to known types from previous field and laboratory studies. Calculated standard error in particle fraction was less than 1 % for all particle types and thus was not included in the discussion. These results are summarized in Table 2.

It is important to describe not only particle sources but also secondary aging impacts as the aging mechanism will change the light absorption cross section of carbonaceous aerosols. For instance, a sulfate coating can increase the light-absorbing properties of soot by a factor of 1.6 (Moffet and Prather, 2009). Internally mixed EC and OC have greater absorption profiles than homogeneously mixed particles of either species (Schnaiter et al., 2005). Additionally, aging can increase particle hygroscopicity through condensation and reaction of gases like NOx or SO2 (Wang et al., 2010; Mochida et al., 2006; Zuberi et al., 2005; Zhang et al., 2008) or oxidation of carbonaceous species (Zuberi et al., 2005; Kotzick et al., 1997). Increased particle hygroscopicity can increase the CCN activity of particles and their growth factor, profoundly impacting radiative effects. Finally, accumulation of nitrogen species on particles can lead to increased deposition of nitrate and ammonium and impact oceanic biology (Paerl, 1995, 1997).

The ATOFMS is a powerful tool with which to measure particle aging because of its ability to measure single-particle composition and directly determine the type and extent of particle aging. For similar particles of the same type, relative peak areas (RPAs) qualitatively reflect the amount of a species on a particle in relation to other species (Bhave et al., 2002; Gross et al., 2000; Prather et al., 2008) and thus can be used to investigate the mechanism of aging (Cahill et al., 2012). During this study, the mixing state of single particles with secondary markers was investigated by identifying and comparing peak areas for ammonium (18NH4+), amines (58C2H5NHCH2+, 59NC3H9+, 86(C2H5)2NCH2+), sulfate (97HSO4-, 195H2SO4HSO4-), nitrate (46NO2-, 62NO3-, 125H(NO3)2-), elemental carbon (12C+, 36C+, 60C+), and organic carbon (27CHN / C2H3+, 29C2H5+, 37C3H+, 43C2H3O+ / CHNO+). Other markers of heterogeneous processing were investigated, but no notable patterns emerged. For this analysis, a particle was considered to be an internal mixed with a species if it had an RPA greater than 0.5 % for the characteristic ion markers, similar to the methodology employed by Cahill et al. (2012).

Size distribution, hygroscopicity, and CCN concentration measurements were collated from periods classified as PGF and CTL. Cumulative CCN supersaturation spectra, defined as median CCN concentration as a function of supersaturation were constructed from the integrated CCN and size distribution data. The spectra were fit to a two-mode hypergeometric model (Cohard et al., 1998) to estimate cloud droplet number concentration (CDNC) for a range of updraft velocity.

The albedo change (ΔAC) of nearshore marine stratocumulus during PGF episodes was determined using Eq. (7) in Platnick and Twomey (1994): ΔAC=AC(1-AC)(χ13-1)AC(χ13-1)+1-1, where χ=NPGF/NCTL is the ratio of CDNC in PGF conditions to CDNC in CTL periods. This analytical formulation relies upon the assumptions of conservative scattering, nearly invariant asymmetry parameter, and constant liquid water path. Furthermore, the estimate we present herein of albedo change during PGF episodes assumes that marine stratocumulus clouds are present near Bodega Bay during PGF and that they are not overtopped by higher cloud layers. The validity of each assumption will be briefly discussed.

Conservative scattering: this assumption is commonly invoked in studies that estimate cloud albedo susceptibility or change (Twomey, 1991; Platnick and Towmey, 1994; Hill and Dobbie, 2008; Hill et al., 2008, 2009; Chen et al., 2011). Liquid cloud particles are generally conservative (single scattering albedo ∼ 1.0) for small to moderate cloud optical depth (τc<23.0). As we will demonstrate, marine cumulus and stratocumulus cloud layers are nearly always below this threshold during PGF.

Invariant asymmetry: for visible light, cloud droplet scattering asymmetry varies weakly with particle size (Kokhanovsky, 2004). For liquid drops, the variation is primarily approximately 5 % over the range of effective radius from 6 to 19 µm. We will demonstrate that the estimated change in liquid drop effective radius during and near PGF periods lies well within this range.

Constant liquid water path: this is the least likely of the above-listed assumptions to be valid. Cloud albedo is susceptible to changes in both cloud droplet number concentration and cloud liquid water path. The latter can also vary with cloud droplet number concentration through cloud dynamic pathways including the so-called “evaporation entrainment” and “sedimentation entrainment” effects (Lu et al., 2005; Wood, 2007; Hill et al., 2009; Chen et al., 2011). The impact of these feedbacks to cloud albedo through the dynamics that control cloud liquid water path vary strongly with environmental conditions and in some cases can cancel the direct increase in cloud albedo resulting from an increase in cloud droplet number concentration. Environmental conditions during PGF (greater likelihood of very dry air above the marine boundary layer, and an increase in large-scale subsidence and thus increased low-level static stability) have been found to favor competing effects on susceptibility through the entrainment effects (e.g., Wood, 2007; Chen et al., 2011). The strength of the entrainment feedbacks is strongly dependent on sea surface temperature as well. PGF can occur under a wide range of sea surface temperatures arising from natural variability in the northeastern Pacific Ocean. To disentangle the total susceptibility which may arise from these competing liquid water path feedbacks, a series of large-scale eddy simulations, similar to those in Lu et al. (2005) and Chen et al. (2011), are required. This is beyond the scope of the current study; thus we will only estimate the so-called “Twomey effect” (or cloud albedo first aerosol indirect effect) on albedo which corresponds to the increase in cloud albedo due to an increase in CCN concentration when liquid water path is held fixed.

The MODIS level 2 cloud products provide swath-level retrievals of liquid cloud optical depth, liquid cloud effective radius, and cloud top pressure twice daily during daylight hours from descending (Terra – 10:15 local time) and ascending (Aqua – 13:45 local time) sun-synchronous orbits. The level 2 cloud products have a nominal spatial resolution of 20 km. For this study, daytime retrievals during PGF conditions from the expanded catalog (N06 PGF events plus Table 3 from this study) during the MODIS operational period (2002–present) were screened to remove pixels over land or more than 75 km from the coast (offshore extent of PGF air mass found by aircraft and reported in N06) and pixels which likely did not correspond to low-level cumulus or stratocumulus. We followed the cloud type definitions (e.g., Fig. 2 from Rossow and Schiffer, 1999) from the International Satellite Cloud Climatology Project (ISCCP) that rely upon thresholds of both cloud top pressure and cloud optical depth. Pixels for which no cloud information was retrieved were also discarded (no cloud present, or retrieval algorithm failed). The retrieved effective radius was also retained to judge the suitability of the invariant asymmetry assumption. The cloud albedo change reported is thus the estimate of the Twomey effect on albedo during PGF episodes when marine cumulus or stratocumulus are present with clear sky above marine low-level clouds.

Table 3 lists all cases which fit the mPGF requirements during CalWater 2015. Hereafter, these will be referred to as PGF(1–5). Some key parameters which describe the PGF layer flow measured by the 449 MHz radar are also summarized in Table 3, along with their ranking among 67 cases (62 cases from N06 plus 5 from CalWater 2015). Note that in all 5 cases both mPGF and elevated CO are met for a majority of the period; however the listed start time and duration in Table 3 is for mPGF, and in some cases the duration for PGF may be shorter than that listed when the additional elevated CO constraint is enforced.

Figure 3 shows a box-and-whisker plot for the peripheral instrument data. So that all measurements fall on the same scale, each measurement has been normalized according to its all-study mean (μall) and has been plotted according to its natural logarithm. For CO, NOx, and CN, the interquartile range during PGF lies entirely above (or nearly so in the case of CN) the interquartile range during CTL. The difference in normalized concentration is most dramatic for NOx, for which the minimum hourly concentration during PGF is nearly the median CTL concentration and the maximum CTL concentration is nearly the median PGF concentration. Median APS number concentration integrated from 542 to 19 810 nm (nAPS) is not preferentially higher during PGF, CTL, or all (ALL) hourly periods, though the range of nAPS for each period varies slightly. PM2.5 is more likely to be elevated during PGF, but its interquartile range overlaps with the interquartile range of PM2.5 during CTL and ALL hourly periods. Mean PM2.5 during ALL hourly periods is estimated to be 14.7±11.6 µg m-3. Mean PM2.5 during CTL periods is estimated to be 14.0±11.9 µg m-3. This estimate increases to 22.9±16.0 µg m-3 during PGF, a mean increase of 64 %.

Figure 3 also displays wind rose diagrams for ALL and CTL periods. The distribution of wind directions and speeds during CTL suggests that these periods are dominated by the land–sea breeze diurnal cycle (BML is situated just east of a shoreline oriented NNW to SSE). The wind rose for PGF is not shown, since wind direction was used in the algorithm for defining PGF.

The range of normalized hourly σATN(λ) values measured during PGF, CTL, and LOCAL periods is shown in Fig. 4. The normalization method follows that used in Fig. 3. As discussed in the Methods section, absolute values are uncorrected and thus not reported. The median and upper/lower quartile values of σATN were compared to published work (Table 1, Chung et al., 2012), and it was found that they are reasonable for the location and concentration/type of aerosols measured. It should be noted that absorption coefficient median and interquartile ranges are highest during PGF, followed by LOCAL and CTL, and that highest maximum values of σATN are observed during LOCAL. The AAE derived from the visible light channels for each period is reported in the figure annotation. The AAE (0.98±0.21) during PGF is very close to 1.0, which is widely accepted to be indicative of fresh soot (Chung et al., 2012). Figure 4 shows that AAE during CTL (0.87±0.10) decreases when compared to PGF. AAE is expected to decrease with soot particle age (accumulation of organic and nitrate on the particle surfaces). A possible explanation for the decrease in CTL AAE in comparison to PGF is that during PGF direct lower tropospheric transport through the Petaluma Gap brings CV and SFBA brings more freshly emitted soot particles to the measurement site. During CTL periods, fewer soot particles are present (lower absorption coefficient), and those that are measured by the aethalometer have been further aged. In Sect. 4.4, this conclusion is supported by single-particle mixing state analysis, which shows that organic : elemental carbon ratio decreases during PGF periods compared to CTL. The AAE during LOCAL periods is highest at 1.17±0.11. This value is closer to that reported for biomass burning (BB; Clarke et al., 2007; Lewis et al., 2008) than is the AAE during PGF or CTL.

Figure 5 shows the average merged size distributions for PGF and CTL sampling periods. CTL periods were marked by lower particle concentrations in the submicron mode and higher particle concentrations in the coarse mode (Dp>1.0 µm). CTL periods often experienced westerly flows and would be expected to be dominated by marine aerosol from the Pacific Ocean. The marine boundary layer over the remote ocean is typified by low particle concentrations and a significant supermicron mode (Quinn et al., 2015). Integrated average supermicron counts on the APS during CTL periods were 14.9 cm-3, compared to 2.4 cm-3 for PGF periods, a decrease of 84 %. Fig. 5b is a log10–log10 plot of Dp and shows supermicron particle concentrations approximately an order of magnitude greater. Single-particle composition data from the ATOFMS during these time periods, discussed more in Sect. 4.3, confirmed that the increase in coarse-mode particles could be attributed to greater concentrations of marine-type particles.

PGF events, in contrast, showed a large increase in the number of particles with Dp<1.0 µm and a new ultrafine mode with mode diameter Dp=36 nm. Figure 5a shows this relationship more clearly, and the corresponding decrease in coarse-mode particles. CN increased by 110 % during PGF compared to CTL, from 311.0 to 650.9 cm-3. These results were also expected, because continental, anthropogenically influenced air masses typically contain smaller and more numerous particles (Tunved et al., 2005). This increase in particle number during PGF events was correlated with the increase in PM2.5 as shown in Fig. 2.

PGF conditions coincide with a shift in particle type away from marine and towards continental. Figure 6 shows pie charts of the sub- and supermicron particle populations for CTL vs. PGF. Percentages indicate the number fraction of particles assigned to the corresponding particle type. The total hit rate for all particles was 20.8 %. Panels a and b show the submicron (0.2–1.0 µm) particle number fraction by type during CTL (1 222 274 particles) and PGF (164 952 particles), respectively. PGF had approximately the same level of submicron BB particles. However, there was a large increase in elemental carbon–organic carbon (ECOC) (20–28 %) and ammonium nitrate (AN) (10–28 %) particle types. These two types have both been linked to anthropogenic activity. The ECOC type has been observed before with the aging of fuel emissions (Hughes et al., 2000) in the Los Angeles Basin during stagnant conditions with high pollutant concentrations. Similarly, the California Central Valley is an area of elevated hydrocarbon fuel emissions and frequent long-lived lower tropospheric inversions, and it might be expected to support the formation of ECOC particles. AN particles have been linked to the accumulation of ammonia and nitric acid on particles (Qin et al., 2012) and nucleation by reactions between ammonia and nitric acid (Russell and Cass, 1986; Hughes et al., 2002), gases strongly correlated with anthropogenic activity. Marine aerosols make up a sizable fration of submicron particles during CTL (SS and aged SS). However, these particles are much less prominent during PGF. The increase in ECOC and AN and decrease in marine particle types reinforce the conclusion that PGF air masses originate continentally and have a strong anthropogenic character.

The clearest delineation in particle type was observed in the supermicron fraction (1.0–3.0 µm). Panels c and d of Fig. 6 show that over 90 % of CTL supermicron particles (544 612 total particles) were either fresh or aged marine particles, while less than 20 % were of marine origin during PGF. During PGF, supermicron particles (25 457 total particles) were primarily comprised of BB; ECOC; AN; and EC, a byproduct of fuel combustion. Additionally, the majority of PGF marine particles showed markers of reacting with nitric acid (Gard et al., 1998). This contrasts CTL marine particles, which were primarily unreacted.

The dust and dust/bio types also increased during PGF. The CV, despite its agricultural production, is a semiarid environment and can be a significant source of dust. Conversely, BML air masses during CTL periods were heavily influenced by the Pacific Ocean and thus were not expected to contain much dust. The shift in supermicron particle composition away from marine particles to anthropogenic and dust particle types supports the conclusion that PGF air masses likely originate from the CV.

Figure 7 shows the sulfate : nitrate ratio (SN) of particles separated by type. Unmodified peak ratios are dependent upon which peak is in the denominator, i.e., whether or not the ratio is greater than 1, thus potentially skewing the data. To account for this, we calculated the normalized peak ratio by the following: ratio > 1, normalized peak ratio = 1-(1/ratio); ratio < 1, normalized peak ratio = ratio - 1; and ratio = 1, normalized peak ratio = 0. As ratios approach -1 or 1, they are exponentially increasing, while close to 1 the RPA of each species is essentially the same. This results in a broader range of ratios for bins near -1 or 1, while the bins near 0 include a smaller range of ratios. The left panel depicts the SN for CTL particles. The majority of particles showed higher SN ratios, indicating that aging primarily occurred through the accumulation of sulfate. The exception to this rule was the SS type, which can react with NOx species in a displacement reaction to liberate HCl (Gard et al., 1998; Cahill et al., 2012). PGF particles showed a SN ratio biased toward nitrate accumulation, indicating that the primary and most important aging mechanism was through NOx pathways.

In addition to probing the partitioning of acidic species, basic species like amines and ammonium were investigated. Figure 8 shows the normalized ratio for amines : ammonium ratio (AA). The AA for CTL particles shows fairly equal partitioning for all particle types. During PGF the average peak area of amine peaks (58C2H5NHCH2+, 59NC3H9+, 86(C2H5)2NCH2+) in particles actually increased, but the AA ratio shifted toward ammonium because the ammonium content of particles increased much more. The area surrounding BML contains animal husbandry but no industrial-scale farming. The shift in basic species partitioning indicates that PGF particles originate from a large source of ammonium. The CV region contains many more industrial-scale farms where ammonium is employed, and thus this change in aging shows that the PGF air mass likely originates within the CV.

Previous studies (Cahill et al., 2012) have used the ATOFMS to determine the internal mixing state of carbonaceous particles. Figure 9 shows the organic carbon : soot ratio as calculated by the ATOFMS, seperated by particle type. CTL particles appear to have relatively higher amounts of OC, most notably in the amine particle types and, unsurprisingly, the OC type. Amines consist of organic carbon chains bound to nitrogen atoms, so it is also unsurprising that these particles would have high OC : EC ratios. The ratio plot indicates that PGF particles contain more EC relative to CTL particles. This is despite the appearance of the AN particle type, which had greater OC character. These OC / EC ratio plots agree with the aethalometer-derived AAE, which suggest that the soot was less aged during PGF than during CTL periods.

In summary this analysis shows that the preeminent aging mechanisms associated with PGF are the accumulation of ammonium and nitrate, in accordance with previous studies on Central Valley particle composition (Qin and Prather, 2006). Amine accumulation was also observed in ECOC and AN particles but was determined to not be as significant as ammonium. Accumulation of nitrogen species on aerosol particles is important as it increases the risk of nitrogen deposition into coastal waters, which can lead to ecosystem degradation (Ryther and Dunstan, 1971; Paerl, 1995, 1997). The shift toward internal mixtures containing elemental carbon and away from particulate matter containing primarily organic carbonaceous species during PGF suggests that gap flow may cause increased solar absorption by near-surface aerosols, especially in visible wavelengths. This potential impact is corroborated by the aethalometer PGF and CTL measurements.

Figure 10 displays the cumulative CCN supersaturation spectrum (versus liquid supersaturation) transformed from the size-resolved CCN data and the CDNC for updraft velocities between 0.1 and 10 m s-1. CCN concentration is enhanced during PGF by 2.8 to 3.0 for a wide range of supersaturations. The increase in CCN is remarkably consistent across the range due to the confluence of two factors. First, the hygroscopicity parameter – or the contribution of particle chemistry to CCN activation – does not change significantly between PGF and CTL (0.21 vs. 0.20). Second, particle concentrations are larger for all sizes D<500 nm during PGF events (Fig. 5). These sizes dominate the spectra in Fig. 10a. As a consequence CDNC increases across all considered cloud updraft speeds during PGF episodes. Figure 10b shows that CDNC increases between 125 and 145 %, and that this relative increase is expected for all possible cloud types. In the results to follow concerning cloud albedo change during PGF, CDNC increases are considered as a ratio (i.e., χ=NPGF/NCTL). In this framework, the average ratio, χ, is 2.35.

During expanded catalog (see Sect. 3) PGF episodes the observed cloud albedo ranged from 0.01 to 0.63, with upper (lower) quartile values of 0.43 (0.17). Using the observed AC=0.43 (0.17) and χ∼2.35, Eq. (7) from Platnick and Twomey (1994) estimates that ΔAC/AC=0.28(0.16). Therefore, clouds that condense on PGF CCN without concomitant changes in liquid water path are expected to be 16 to 28 % more reflective when considering the Twomey aerosol indirect effect. As discussed in the Methods section, the values reported here correspond to PGF episodes where marine cumulus or stratocumulus clouds are present within 75 km of the shoreline, and, when clear, the marine cloud layer is topped by clear sky.