Observations of heterogeneous reactions between Asian pollution and mineral dust over the Eastern North Paciﬁc during INTEX-B

. In-situ airborne measurements of trace gases, aerosol size distributions, chemistry and optical properties

their optical properties compared to their variability in-situ. However, competition between supermicrometer dust and submicrometer primary aerosol for condensing secondary aerosol species led to a 25% smaller number median diameter for the accumulation mode aerosol. A 10-25% reduction of accumulation mode number median diameter results in a 30-70% reduction in submicrometer light scattering at relative humidities in the 80-95% range. At 80% RH submicrometer light scattering is only reduced ∼3% due to a higher mass fraction of hydrophobic refractory components in the dust-affected accumulation mode aerosol. Thus reducing the geometric mean diameter of the submicrometer aerosol has a much larger effect on aerosol optical properties than changes to the hygroscopic:hydrophobic mass fractions of the accumulation mode aerosol.
In the presence of dust, nitric acid concentrations are reduced to <50% of total nitrate (nitric acid plus particulate nitrate). NO y as a fraction of total nitrogen (NO y plus particulate nitrate), is reduced from >85% to 60-80% in the presence of dust. These observations support previous model studies which predict irreversible sequestration of reactive nitrogen species through heterogeneous reactions with min-

Introduction
Mineral aerosol is generated at the Earth surface by aeolian erosion of unconsolidated sand to clay grade soil particles. Under specific meteorological conditions, large dust storms can loft mineral dust directly into the FT where they can be transported intercontinentally (Clarke et al., 2001;Husar et al., 2001;Prospero, 1999).
Estimates of present day dust emissions fluxes are on the order of 1000-3000 Tg yr −1 (Dentener et al., 1996Ginoux et al., 2001). The total all-models-average for the global climate models (GCMs) participating in the Aero-Com project is 1840 Tg yr −1 with a total model diversity 1 , δ, of 49% . The AeroCom GCMs estimate that wet removal accounts for 32% of total removal, though there is considerable variability between the models (δ = 54%). The mean atmospheric residence times are on the order of 4 days (δ = 43%) resulting in a mean estimate of global annual average aerosol dust burden of 20 Tg (δ = 40%). AeroCom simulations of mineral dust indicate this aerosol accounts for ∼70% of the global annual average dry aerosol mass. However, dust accounts for only 25% of global annually average aerosol optical depth (AOT), a value comparable in magnitude to hygroscopic aerosol such as sulfates and sea salt .
The Chinese "Loess plateau" is the world's largest deposit of loess; a sedimentary deposit of wind blown siltand clay-grade material. The loess-paleosol sequence of the Loess plateau has a relatively continuous depositional character over the past 2.4-2.6 million years (Liu, 1985). Re-cent work potentially extends the geological time scale of its formation to ∼7.0 million years (Ding et al., 2001). Electron microscope analysis of Chinese Loess show that the upper part of the stratigraphy is comprised of a meta-stable microfabric of plate-like aliminosilicates (∼>70 wt%) (Derbyshire, 1983;Ding et al., 2001;Liu, 1985). Aluminosilicates in the clay size fraction (<6.0 µm) of Chinese Loess are composed mostly of illite, kaolinite and montmorillinite (Liu, 1985). Calcite is the fourth most abundant mineral in the clay size fraction (<6 µm) but is also the most variable. During suspension Ca/Al ratios in Asian dust are depleted compared to the parent materials while Fe/Al ratios are relatively well conserved (Arimoto et al., 2006 and refs). Iron oxides constitute ∼3-6 wt%, and are typically surface bound to the clay grade minerals accounting for ∼50% of the total iron (Fe 2 O 3 ). Goethite and hematite account for 75% and 25% (+/−10%) of the iron oxides respectively and are the only two minerals responsible for visible light absorption by Asian dust (Lafon et al., 2006). Laboratory tests of water adsorption indicate no more than 3-4 monolayers of H 2 O on the surface of these minerals under typical atmospheric conditions (Grassian, 2001 and refs). HNO 3 and NO 2 will adsorb directly onto metal oxides (Al 2 O 3 , Fe 2 O 3 , TiO 2 , CaO and MgO) with the notable exception of SiO 2 . The presence of adsorbed water greatly enhances the NO 2 and HNO 3 uptake coefficients compared to dry conditions. However reactions between NO 2 and HNO 3 with Al 2 O 3 , Fe 2 O 3 and TiO 2 are surface limited, i.e. there is no reaction with internal crystal structure, whereas the reactions with CaO, MgO, and especially CaCO 3 , occur throughout the bulk particle (Grassian, 2001).
The recognition that anthropogenic secondary aerosol precursors (e.g. SO 2 , NO 2 and HNO 3 ) can react with natural mineral aerosol, has led to several model studies designed to evaluate the effects of heterogeneous chemical reactions on aerosol direct and indirect effects (Bauer et al., 2007;Fan et al., 2004;Tang et al., 2004). Heterogeneous reactions were found to increase wet deposition of dust near the Asian source regions resulting in up to a 50% decrease in deposition over the Eastern North Pacific (Fan et al., 2004). Citing Lammel and Novakov (1995) as well as Wyslouzil et al. (1994), Bauer et al. (2007) assumed dust particles with a 10% surface coating of sulfate or nitrate behaved as if they were completely soluble. The studies cited are hardly applicable to the formation of "soluble dust", as they investigated the coating of homogeneous, hydrophobic, primary soot with soluble sulfate and nitrate species. However, based on this assumption, Bauer et al. (2007) conclude that enhanced wet deposition of "soluble dust" leads to a 20% reduction (33.5 vs. 41.6 Tg) in the present-day global annual dust budget, compared to simulations that omit heterogeneous reactions.
Particles with diameters greater than ∼2 µm activate regardless of composition for supersaturations typical of continental and marine cumulus cloud (∼>0.2%) (Kelly et al., 2007). At 0.2% supersaturation CaCO 3 and SiO 2 with a 1% coating of gypsum, will activate if they have dry diameters greater than ∼1 µm. Activation of "completely insoluble" dust particles in the 0.6-2.0 µm size range is facilitated by the presence of slightly soluble compounds (Kelly et al., 2007). These findings are supported by field measurements which show that calcite-containing particulate are acting as CCN (Matsuki et al., 2009;Twohy et al., 2009). The size, composition and mixing state of dust particles will codetermine their effectiveness as ice nuclei (IN). At −10 • C to −40 • C and relative humidity with respect to ice (RH i ) of 110-135%, kaolinite, illite, muscovite and hematite are the most efficient IN while montmorillonite, quartz, calcite and gypsum are the poorest IN (Eastwood et al., 2008;Zimmermann et al., 2008). Eastwood et al. (2009) recently determined that heterogeneous reactions can inhibit the icenucleating ability of kaolinite. Whereas a separate study (Archuleta et al., 2005) found that Asian dust particles larger than ∼0.2 µm are all effective cirrus cloud ice nuclei at RH i of ∼135% irrespective of temperature.
Given the difficulty in simulating relative humidity (Petch, 2001) and aerosol indirect effects using GCM's , it is difficult to have confidence in recent model estimates of the influence of heterogeneous chemical reactions on atmospheric residence times of mineral dust. In their 2007 work, Bauer et al. (2007) calculate a 20% reduction in present-day dust burden compared to pre-industrial values (33.5 vs. 41.8 Tg) when heterogeneous chemical reactions are considered in both simulations. Present-day versus pre-industrial simulations should potentially include emissions of primary urban dust (Alfaro et al., 2003;Cohen et al., 2004), soil dust from agricultural activities (Sokolik and Toon, 1996;Tegen and Fung, 1995;Tegen et al., 2004), and dust emissions due to anthropogenically-induced desertification (Chen et al., 1999;Moulin and Chiapello, 2006;Sheehy, 1992). Furthermore, short model integration times (e.g. 6 years in Bauer et al.) cannot capture inter-decadal variability in precipitation and vegetative cover; which likely exerts a controlling influence on dust generation (Dai et al., 1997;Nicholson et al., 1998). The absence of dynamic aerosol size distributions with size resolved, and source-dependent, mineral compositions in GCM's mean that recent comparisons between pre-industrial versus contemporary dust burdens are likely an over-simplification.
Here we use in-situ airborne measurements of trace gases, aerosol microphysics and chemistry to evaluate the potential for heterogeneous chemical reactions between Asian pollution and Asian dust to influence aerosol optical properties over the Eastern North Pacific. We include an analysis of the heavily polluted continental boundary layer (CBL) and free troposphere (FT) near Mexico City, as well as longterm measurements at the Mauna Loa Observatory (MLO) in Hawaii. These results are then discussed with regard to modeling studies in order to evaluate model assumptions and the plausibility of some of their conclusions regarding heterogeneous reactions occurring on mineral dust particulate, their effect on aerosol optical properties, and the partitioning of reactive species between the gas and aerosol phases.

In-situ measurements of aerosol size distributions
Aboard the NASA DC-8 the University of Hawaii solid diffuser inlet (UH inlet) delivers ambient air to the combined NASA Langley Research Center (LaRC) and Hawaii Group for Environmental Aerosol Research (HiGEAR) aerosol sampling package. Sample air is delivered to the University of New Hampshire's (UNH) filter-based aerosol chemistry system via the UNH solid diffuser inlet. Supermicrometer aerosol passing efficiencies were recently evaluated for these two inlets over the performance envelope of the NASA DC-8 (McNaughton et al., 2007). Aboard the NSF/NCAR C-130 sample air is delivered to the HiGEAR instrument package via a solid diffuser inlet evaluated during PELTI .
Total and refractory (residence time, τ = 0.1 s @ 360 • C) aerosol number (Dp>10 nm) are measured using two TSI model 3010 condensation nuclei counters (Clarke et al., 1997). Custom built differential mobility analyzers (DMA) are used to measure aerosol size distributions over the d m = 0.010-0.20 µm size range aboard both aircraft. The DMAs are equipped with a lagged aerosol grab sampler (LAG chamber) (Clarke et al., 1998). Each DMA system is also equipped with a heater assembly or, thermo-optical aerosol discriminator (TOAD), which pre-heats the aerosol to 150 • C or 300 • C (τ = 0.2 s) prior to analysis (Clarke, 1991). In polluted airmasses the internally mixed refractory aerosol is commonly comprised of the "soot" components responsible for light absorption Mayol-Bracero et al., 2002). Preheating the aerosol does not affect the analysis of sea salt or dust, the most common natural refractory primary aerosol species. During INTEX-B the DC-8 aerosol package was equipped with an additional "long" DMA (LDMA, d m = 0.01-0.50 µm) and employed a smaller mini-LAG chamber but no TOAD. The MILAGRO C-130 did not contain an additional LDMA unit.
Both aircraft were equipped with a custom modified PMS LAS-X optical particle counter used to measure the aerosol size distribution between d oe = 0.1 and 20.0 µm at a size resolution of 112 channels per logarithmic decade. Each OPC is equipped with a 4-channel TOAD assembly operating at dry 2 ambient temperature, 150 • C, 300 • C and 420 • C (τ = 2.0 s) . The OPC's are calibrated using monodisperse polystyrene spheres with a density of 1.05 g cm −3 and a refractive index of 1.59 @ 589 nm. OPC sizing accuracy is also evaluated using borosilicate glass beads with a density of 2.52 g cm −3 and a refractive index of 1.56 @ 589 nm, and silicon dioxide (SiO 2 ) spheres with a density of 2.01 g cm −3 and a refractive index of 1.40 @ 589 nm. The aerosol size determined by the OPC instrument are optically effective diameters (d oe ) as discussed in Clarke et al. (2004). During post-processing the unheated OPC size distributions are adjusted to d g , to account for sizing errors due to aerosol refractive indices that differ from those of the PSL calibration spheres (1.59 @ 589 nm). OPC sizes over the 0.12-0.53 µm size range are adjusted assuming an ammonium sulfate composition with a refractive index of 1.53-0.0i. When sampling mineral dust the OPC sizes over the size range 0.53-8.0 µm are adjusted assuming a refractive index of 1.53-0.0006i. When sampling sea salt aerosol no size adjustment is made to the data in the 0.53-8.0 µm size range because the dry sea salt refractive index (1.588-0.0i) is close to that of PSL (1.59-0.0i) at the He-Ne laser wavelength of 633 nm. No optical to geometric size adjustments are made for the heated OPC channels as information regarding chemical composition, and thus refractive index, is a relative unknown. Counting statistics and sizing accuracy of the OPC is typically poor beyond 8.0 µm, thus data above this size range is typically eliminated from the data sets.
Aerodynamic aerosol size distributions in the d ae = 0.5-20.0 µm size range are measured using a TSI model 3321 aerodynamic particle sizer (APS). APS flow and sizing calibrations were routinely performed according to the procedures outlined in McNaughton et al. (2007). Data from 2 ram heating plus 50% dilution with desiccated air. the first five channels (0.50-0.78 µm) were discarded due to poor instrument performance over this size range. Ignoring slip correction factors (C c,ae /C c,g ) by approximating them as unity, aerodynamic diameters (d ae ) were adjusted to geometric diameters (d g ) during post processing (Baron and Willeke, 2001;DeCarlo et al., 2004). We assume a dry bulk density of 2.06 g cm −3 for Chinese loess (Liu, 1985). This bulk density is used to correct supermicrometer mineral aerosol aerodynamic to geometric diameters and when converting aerosol volume to mass and vice versa. This is equivalent to assigning a shape factor, χ, of 1.10-1.25 for dust particles with aerodynamic diameters between 0.5 and 10.0 µm assuming a bulk density for crustal material of 2.56 g cm −3 (Craig., 1997). The loess value (2.06 g cm −3 ) is nearly identical to the "effective density" of 2.0 g cm −3 proposed by Reid et al. (2003) but smaller than the value used in McNaughton et al. (2007) to evaluate DC-8 inlet passing efficiencies. When recalculated, the 50% passing efficiency diameters of the UH and UNH inlets are no less than 3.5 µm and 2.8 µm when sampling Asian dust at the surface and 2.2 µm and 1.8 µm when sampling at the DC-8 ceiling of 12 km.

In-situ measurements of aerosol optical properties
Total aerosol light scattering is measured aboard each aircraft using TSI model 3563, 3-wavelength (3-λ) integrating nephelometers (TSI-Neph) (Anderson et al., 1996;Anderson and Ogren, 1998). Submicrometer aerosol light scattering is measured using Radiance Research model 903 singlewavelength nephelometers (RR-Neph). The Anderson and Ogren (1998) truncation correction has been applied to the TSI-Neph data while the empirically derived truncation correction of Anderson et al. (2003) has been applied to the RR Neph data. Calibrations using filtered CO 2 and Refrigerant 134A were performed prior to each deployment and every 3-5 flights. No significant (<5%) calibration adjustments were required for the TSI-Nephs while slight (5-10%) adjustments were often required for the zero and span of the RR-Nephs. During each flight at least one 300-s filtered air sample was collected to check for leaks and/or deviations from zero. Accuracy (+/−2%) and precision (0.2 Mm −1 for 300s average) estimates for these instruments are comparable to those determined previously (Anderson et al., 2003;Anderson and Ogren, 1998).
Total and submicrometer aerosol absorption is measured using Radiance Research 3-λ particle soot absorption photometers (PSAP) corrected according to Virkkula et al. (2005). In laboratory tests prior to the MILAGRO and INTEX-B experiment three of the four 3-λ PSAPs were tested for instrument noise while sampling filtered air. The average instrument noise (δ 300 ) is computed as the average standard deviation for six, 300 s (5-min) averages of the 1-Hz data (Table 1). These values are comparable to instrument noise (δ 240 = 0.72 Mm −1 and 0.33 Mm −1 ) reported for Two RR-Nephs are used to measure aerosol f(RH) between 80% and 40% RH at a wavelength of 532 nm. Air sampled by one nephelometer (RRwet) is preconditioned to a relative humidity of 80% +/−7% using a custom designed f(RH) system employing a GORE-TEX T M membrane. The second nephelometer (RRdry) was plumbed in parallel and outfitted with a commercial blanket heater. The blanket heater raises the nephelometer temperature by 5-10 • C reducing the relative humidity to values below 40% in the humid (RH amb >90%) marine boundary layer and often below 5% in the free troposphere. This heating may cause minor evaporation of non-refractory species of the order of a several percent of the mass of organic species and ammonium nitrate (Huffman et al., 2008). Using the corrected scattering values and RH recorded by the RRdry and RRwet nephelometers, a two point fit is used to compute γ , the exponential term in the f(RH) equation: where, σ sp,dry and σ sp,amb are light scattering at the indicated "dry" and "ambient" relative humidities (Carrico et al., 2003;Howell et al., 2006). A ground-based test of the f(RH) system was performed in its flight configuration aboard the NASA DC-8. The results indicate that scattering measurements for ammonium sulfate and sea salt (from a filtered North Pacific subtropical gyre seawater standard; Karl and Lukas, 1996) are within 25% of the values calculated from size distributions considering salt densities and optical properties at 80% relative humidity (Tang, 1997;Tang and Munkelwitz, 1994;Tang et al., 1997). Since f(RH) is computed from a two point fit of the ratio between the dry and wet scattering values the f(RH) measurement could be low by up to a factor of 0.7 or overestimated by up to a factor of 1.3.

Fitting aerosol size distributions using lognormal distributions
In this analysis, DMA, OPC and APS size distributions are averaged, combined, then fit using log-normal distributions (Baron and Willeke, 2001;Seinfeld and Pandis, 1998). The largest disparity occurs between the OPC and the LDMA over the 0.30-0.50 µm size range. OPC number agreement is 71% (R 2 =0.922) and volume agreement is 60% (R 2 = 0.910) due to the large OPC Mie oscillation over this size range. The overall agreement for the OPC and LDMA between 0.12 and 0.50 µm is 92% (R 2 = 0.918) for aerosol number, and 85% (R 2 = 0.966) for aerosol volume. Other researchers have applied log-normal fitting routines using least squares optimization (Hand and Kreidenweis, 2002;Osborne and Haywood, 2005;Porter and Clarke, 1997). However, in the cases cited only one moment of the aerosol size distribution is considered during the fit. When fitting the number distribution (Osborne and Haywood, 2005;Porter and Clarke, 1997) the tail of the distribution does not significantly contribute to the least squares error. But, the tail of the number distribution contains all of the aerosol volume and will be poorly constrained by fitting the distribution using only aerosol number. Similarly, when fitting using the volume distribution (Hand and Kreidenweis, 2002) aerosol number will be poorly fit.
To avoid the problems associated with fitting a single mode of the distribution, we (S. Howell) developed a least squares fitting routine which simultaneously evaluates the least squares error for the number, length, area and volume distributions (i.e. the zeroth through third moments). This technique is similar to that of Stroud et al. (2007), who simultaneously fit the number and volume distribution from an SMPS and the mass distributions from an aerosol mass spectrometer. Using multiple moments results in better fits, which do not bias the results by large errors contributed from a single mode. Typically, one to three modes (termed the C. S. McNaughton et al.: Observations of heterogeneous reactions between Asian pollution and mineral dust Aitken, accumulation and coarse modes), were required to fit the unheated distributions. In general only two fits, Aikten and coarse mode, are required to fit the refractory aerosol distributions. As indicated in our previous studies , the refractory Aitken mode aerosol is found to be an internally mixed component of some fraction of the unheated accumulation mode aerosol.

Trace gas measurements
DC-8 trace gas measurements of NO 2 , HNO 3 , total peroxynitrates (PNs), alkyl-and hydroxyalkyl nitrates (ANs) were measured by the University of California, Berkeley using thermal-dissociation coupled to laser-induced fluorescence (TD-LIF) (Day et al., 2002;Thornton et al., 2000). Based on previous studies (Miyazaki et al., 2005;Zondlo et al., 2003) and recent measurements during ARCTAS (P. Wennberg, personal communication, 2009), it is likely that the UC Berkeley TD-LIF and UNH DC-8 mist chamber measurements of HNO 3 include a 10-30% enhancement due to contamination by NH 4 NO 3 aerosol. Larger enhancements are possible in the presence of dust (Miyazaki et al., 2005). For INTEX-B the two HNO 3 measurements are highly correlated over more than 2 orders of magnitude (m = 0.89, R 2 = 0.83) and are averaged to produce a single measure of HNO 3 . No attempt has been made to correct the HNO 3 data for particulate nitrate contamination. SO 2 was measured by the Georgia Institute of Technology chemical ionization mass spectrometer (Kim et al., 2007). NO was measured using chemiluminescence (Sjostedt et al., 2007). HCl was measured by the Georgia Institute of Technology using a chemical ionization mass spectrometer (Kim et al., 2008 and refs). Ozone was measured using nitric oxide chemiluminescence by the NASA Langley Research Center (Davis et al., 2003).
Aboard the NSF/NCAR C-130, HNO 3 was measured by the California Institute of Technology using a chemical ion-ization mass spectrometer (Crounse et al., 2006). NO, NO 2 , NO y , and O 3 were measured by a team from the National Center for Atmospheric Research (NCAR) using a 1-Hz chemiluminescence technique (Weinheimer et al., 1998). PAN was measured by an NCAR team using thermal decomposition chemical ionization spectrometry (Slusher et al., 2004).  (Dunlea, 2009;Singh et al., 2009). After excluding clean FT airmasses (CO<90 ppbv; O 3 <40 ppbv), and those indicative of stratospheric influence ( 7 Be>800 fCi sm −3 , where sm −3 denotes concentrations per cubic meter of air at 1013.25 mb and 273.15 K), the data were further stratified using the fine mode fraction of aerosol light scattering (Anderson et al., 2003).

Anthropogenic pollution and Asian Dust over the Eastern North Pacific
When FMF scat >0.6, and scattering exceeded 3 Mm −1 , the airmasses are considered "fine" mode dominated anthropogenic pollution. Values of 0.3 < FMF scat <0.6 are labeled "mixed" pollution and mineral dust. No data met the "dusty" FMF scat <0.3 criteria of Anderson et al. (2003).
Lognormal fits of the anthropogenic and mixed aerosol size distributions measured near Hawaii are presented in Fig. 1. The top two panels are the number and volume distributions for the fine mode dominated case whereas the bottom panels are for the mixed airmass type. Differences between the unheated and refractory fits to the supermicrometer dust mode are not significant and the summary statistics in Table 2 are pooled into a single supermicrometer aerosol fit. Figure 2 assesses the degree of aerosol number closure by comparing the integrals of the lognormal fits to the average concentrations measured using the hot and cold CN counters (N for Dp>10 nm sampled at 1-Hz). Size distribution integrals reasonably replicate the faster CN counter averages and support the use of these fits to constrain model simulations of the aerosol size distribution. Figures 3 and 4 are the fits and the number closure assessment for the data collected in the FT of the Aleutian Low. For these cases refractory number        is somewhat higher in the fit data resulting in fair agreement between the RCN ratios. Table 2 summarizes the mean log-normal fits to the accumulation mode, coarse mode and refractory Aitken mode size distributions measured near Hawaii and Alaska for the fine mode dominated and mixed airmass types. After long range transport, the number median diameter (NMD) of FT Asian pollution (FMF scat >0.6) aerosol varies (+/−1-σ ) between 0.086 and 0.15 µm with an average value of 0.11 µm and a geometric standard deviation, σ g , of 1.70+/−0.15. The observed range (+/-1-σ ) of volume median diameters (VMD) is between 0.24 and 0.30 µm. In the presence of dust (0.3<FMF scat <0.6), the NMD of FT Asian aerosol varies between 0.051-0.11 µm with a smaller average NMD of 0.075 µm. In the presence of dust the accumulation mode aerosol has a broader distribution, σ g = 1.97+/−0.20, which results in a range of VMD's between 0.26 and 0.35 µm. Despite their larger VMD's, for a fixed number of particles, the average value of the fit to the "mixed" aerosol types result in 30% less aerosol volume than under low-dust conditions. This is close to the significant difference (α = 0.05) of 40% between the average accumulation mode volumes for the observed distributions (V anthro = 1.4 +/− 1.5 µm 3 cm −3 ; V dusty = 0.86 +/−0.64 µm 3 cm −3 ); a quantity that will vary due to factors such as emissions source strength, transport pathway etc.
The total and refractory number distributions highlight the internally mixed nature of the refractory soot particulate coated with the non-refractory compounds, NH + 4 , SO 2− 4 , NO − 3 and organics. The ratio of refractory number to total number in these airmasses averages 0.75 indicating only 25% of the pollution aerosol is composed of externally mixed secondary aerosol number. The refractory aerosol NMD's for Hawaii and Alaska vary between 0.032 and 0.066 µm with a mean value of 0.046 µm. The geometric standard deviation is variable but averages 1.91+/−0.23. This results in a range of VMD's between 0.10 and 0.26 µm for the refractory material. The average refractory volume ratio of the log normal fits is 14-19%, in good agreement with the observed average refractory volume ratios of 12-16%+/−7%. The INTEX-B observations of reductions in accumulation mode secondary aerosol mass due to competition by dust surface area is consistent with previous observations near the Asian source region (Howell et al., 2006). Models which attempt to simulate these aerosol should account for this internal mixing as surface coatings of non-absorbing secondary aerosol on internally mixed carbonaceous primary aerosol will affect aerosol optical properties Fuller et al., 1999;Schnaiter et al., 2005).
Total aerosol single scatter albedo (SSA) at 550 nm measured 0.94+/−0.03 and 0.95+/−0.01 for fine mode dominated (FMF scat >0.6) airmasses sampled near Hawaii and Alaska. Using a one-tailed Student's t-test these values are Table 3. Summary of total single scatter albedo, scattering and absorptionÅngstrom exponents for 300-s level leg averages during INTEX-B. "Anthro" are fine mode dominated aerosol (FMFscat >0.6) while "Mixed" aerosols are for mixtures of pollution and dust (0.3< FMFscat <0.6). The standard deviations of the means are shown in brackets.   Table 3 along with the values measured in the continental boundary layer over the Central Mexican Plateau and the FT and MBL over the Gulf of Mexico. Measurements of 3-λ total aerosol scattering and absorption allow us to investigate the wavelength dependence of extinction for Asian aerosol after long-range transport to the remote North Pacific. Table 3 summarizes the total scattering Angstrom exponent as well as the total absorptionÅngstrom exponents measured for the fine mode dominated and mixed aerosols. The averages are computed from 5-min level leg FT averages where the measured absorption is above 0.1 Mm −1 . Values from the Central Mexican Plateau and the FT over the Gulf of Mexico are also included for comparison. The Hawaii data show the lowest scatteringÅngstrom exponents and the highest absorptionÅngstrom exponents while the data from the CBL near Mexico City have the lowest singlescatter albedo's.
Submicrometer absorptionÅngstrom exponents are always higher when dust is present (not shown). The f(RH) measurements (Table 4) also show that submicrometer f(RH) γ is suppressed (0.39-0.44) in the presence of dust when These two measurements indicate that a portion of the dust is passing the 1.0 µm aerodynamic (at STP) cut-off of the impactor plate. Ignoring shape factor, the impactor's 50% cut corresponds to a geometric diameter of 0.67 µm for a mineral dust aerosol with a density of 2.06 g cm −3 . Based on the lognormal fit for the Hawaii mixed aerosol data (N = 5.3 cm −3 , NMD = 0.64 µm, σ g = 2.06), this "submicrometer" dust is 0.3 µg m −3 of the total, 15.7 µg m −3 . Using a refractive index of 1.53-0.0006i, the "submicrometer" tail of the dust distribution, which accounts for just 2% of the dust mass, accounts for 11% of the extinction by mineral dust.  Poorly modeling the lower tail of the mineral aerosol size distribution (2% of dust mass) can result in larger errors in aerosol extinction (11% of dust extinction at 550 nm). As shown by the wavelength dependence measurements, this error will compound when calculating extinction at shorter visible wavelengths and in the UV (Sokolik and Toon, 1999) as well as for dusts with differing composition (e.g. Lafon et al., 2006). Thus accurately modeling the direct and indirect effects of mineral dust likely requires dynamic aerosol size distributions with size resolved, and source-dependent, mineral composition.

Heterogeneous chemistry of Asian Dust over the Eastern North Pacific
The previous section demonstrated that Asian pollution and dust commonly occur together in polluted airmasses transported long-range across the Pacific. In this subsection we examine the chemistry of accumulation mode and coarse mode aerosol measured in the pollution/dust plumes over the Eastern North Pacific. University of New Hampshire filterbased measurements of Ca 2+ can be converted into calcite (CaCO 3 ) mass assuming that the weak-acid eluent of the UNH ion chromatography analysis does not measure the aluminosilicate's structural calcium. When these values are regressed against the HiGEAR data (Fig. 5) we estimate a calcite content of 11 wt% for the Hawaii region (R 2 = 0.90) and 10 wt% (R 2 = 0.57) for the Alaska region. The low R 2 of the Alaskan fit is being driven by the two highest values of aerosol Ca 2+ . A calcite weight percent of 10-11% is in good agreement with soil carbonate values reported by Liu (1985) (3.6-21%; mean value of 12%), and 14-17% (d m <40.0 µm) reported by Wang et al. (2005). The 10-11% derived here for aerosol carbonate is also relatively consistent with values reported for clay grade (d m <6.0 µm; 6.5+/−6.2%) Asian particulate (Liu, 1985), measurements of 5-8 wt% soluble aerosol calcium during ACE-Asia (Kline et al., 2004) and 8.2+/−3.2 wt% measured in aerosol collected by Wang et al. (2005). As per Wang et al. (2005), the high calcite content measured during INTEX-B is indicative of a Western Chinese source provenance for this material.
The ratio of sulfate aerosol to the sum of gas phase SO 2 plus sulfate aerosol (S-ratio) has been shown to increase with airmass age (Brock et al., 2008;Dunlea, 2009). Values greater than ∼0.5 generally indicate airmass ages greater than ∼3 days, although lower values can be observed in older airmasses after wet deposition. In the presence of the dust we can expect heterogeneous reactions to drive the S-ratio toward unity at a faster rate. Jordan et al. (2003a) evaluated the ratio of particulate nitrate (NO − 3 ) to nitrate plus nitric acid (N-ratio) in order to evaluate the heterogeneous uptake of NO x species by alkaline dust. Their observations show that in the presence of dust, particulate nitrate accounts for 50-60% of total nitrate. The values range between 25% and 50% without dust. Using single particle analysis Sullivan et al. (2007a), found that Fe-rich Asian dust particles were closely associated with secondary sulfate whereas Carich particulate could contain secondary sulfate, nitrate and chloride. Reaction between aerosol carbonate and HCl derived from the acidification of seasalt by SO x or NO x has also recently been observed (Murphy et al., 2006;Sullivan et al., 2007b). Figure 6 plots the S-ratio, N-ratio, a modified N-ratio, and the ratio of aerosol chloride to HCl plus chloride (Cl-ratio) versus calcium as measured between Hawaii and Alaska during INTEX-B. The plots are color coded by scattering Angstrom exponent. High values (red = 3.0) are indicative of small particles found in pollution dominated airmasses. Low values (blue = −0.5) are indicative of airmasses whose optical properties are dominated by supermicrometer dust. For the S-ratio (far-left) there is no clear dependence with calcium concentration. This indicates that airmass age and/or scavenging could control the partitioning of sulfur. The Nratio versus Ca 2+ plot (center-left) shows that for low Ca 2+ concentrations (< 20 neq sm-3) the ratio averages 47% (+/-22%) and that at higher concentrations (>20 neq sm-3) the value averages 65% (+/−25%), a significant difference using a two-tailed Student's t-test (α = 0.05). The effect of these reactions is even more striking when aerosol nitrate is divided by the sum of all nitrogen species (NO,NO 2   is clearly sequestering gas-phase nitrogen species into the particulate-phase (Fig. 6, center-right). Aerosol nitrate accounts for only 10% (+/−6%) of the nitrate in the low dust group (<20 neq sm-3 Ca 2+ ) whereas it averages 26% (+/−11%) in the high dust group and has maximum values approaching 50%. A similar trend is evident for the uptake of Cl − by dust. We find that 88% (+/−7.5%) of the chloride is associated with particulate for low dust concentrations (Ca 2+ <20 neq sm-3) compared to 92% (+/−3.6%) for high dust concentrations; values which are significantly different (α = 0.05, p = 0.02). These reactions are explored further in Fig. 7. The top row plots aerosol SO 2− 4 , NO − 3 and Cl − in neq sm −3 (T=273.15 K, P=101.325 kPa) versus the equivalents of Ca 2+ . The regression for total sulfate is poor and does not improve after subtracting the submicrometer sulfate. And although Chinese Loess is not rich in CaSO 4 (Liu, 1985), Arimoto et al. (2004) reported an SO 2− 4 to WS Ca 2+ molar ratio of 0.09+/−0.04 for clean, dust dominated (>40 µg m −3 ) aerosols at Zhenbeitai, China during ACE-Asia. A ratio of 0.1 is also the minimum value measured by Kline et al. (2004) aboard the NSF/NCAR C-130 during ACE-Asia. The poor fit (R 2 ∼ = 0.20) and slope of 0.15 for supermicrometer only aerosol means we cannot rule out the possible presence of gypsum. The regression for nitrate indicates a close coupling between the presence of aerosol nitrate and mineral dust (m = 0.14, R 2 = 0.69). Primary Ca(NO 3 ) 2 can likely be ruled out as Arimoto et al. (2004), reported a molar equivalents ratio of NO − 3 /WS Ca 2+ of just 0.008. The presence of chloride as a function of calcium is not robust (R 2 = 0.05).
When the degree of aerosol neutralization is plotted against equivalents of calcium (middle-row of Fig. 7), and the data are coloured by scatteringÅngstrom exponent, there is generally sufficient ammonium to neutralize most SO 2− 4 to form (NH 4 ) 2 SO 4 . Also, ammonium is generally insufficient to create much NH 4 NO 3 nor is it available to combine with Cl − . Using the method of Jordan Jordan et al. (2003a) regressions were for near source Ca 2+ values that ranged from 100-1000 neq sm −3 whereas the INTEX-B concentrations are substantially reduced (max ∼130 neq sm −3 ) due to removal processes (Dunlea, 2009), and dilution during long-range transport. The regression for the sum of excess sulfate plus nitrate versus calcium is more significant (R 2 = 0.35) than that for sulfate alone (R 2 = 0.12) but each of the regressions which underweight outliers are similarly robust. While there is the potential that sulfate, nitrate and calcium are simply co-emitted at the source, Sullivan et al. (2007a) found ammonium can be associated with sulfate rich submicrometer dust leading to inaccuracy in the Jordan et al. (2003a), assumption that all ammonium is associated with submicrometer sulfate or nitrate. Furthermore, when combined with the potential influence of submicrometer calciumcontaining particulate of anthropogenic origin (e.g. fly ash), the inherent uncertainty in airborne measurements of supermicrometer aerosol (Blomquist et al., 2001;Huebert et al., 2004;McNaughton et al., 2007) as well as ammonium nitrate (Bergin et al., 1997), it is not surprising that these regressions are not very robust. Still, the trends indicate that up to 13% of the calcite alkalinity could have been consumed by acidic sulfur species, and that ∼14% of the calcite alkalinity has been consumed by acidic nitrogen species. In the absence of dust these secondary aerosol species may have remained in the gas phase or condensed upon the submicrometer particles. When chloride is considered, the modified R 2 statistic increases slightly, demonstrating that a further 4% consumption of the calcite alkalinity could be related to heterogeneous reactions between dust and HCl.   For irreversible reactions with calcite to account for the increases in aerosol sulfate, nitrate and chloride we need to establish whether or not the Ca 2+ measured in the Asian plumes is actually calcite. We have already used the size distribution measurements to derive an 11 wt% of CaCO 3 for the dust aerosol (Fig. 5). The variability of the calcite wt% in Chinese Loess (3.6-21%, Liu, 1985) leaves a relatively large margin for error and the 11 wt% reported here is higher than the near source values (5-8 wt%) reported by Kline et al. (2004). The left panel of Fig. 8 plots the anions versus cations for the UNH filter-based chemistry measurements. Minor cations such as Fe 2+ are not measured by the UNH IC technique but their concentrations are expected to be a small fraction of the total. Similarly H + is not reported thus the pH of the aerosol is unknown. The major missing anion in the analysis is CO 2− 3 . Where a large anion/cation imbalance occurs we can assess the probability that the missing ion is CO 2− 3 by plotting the imbalance vs. Ca 2+ (center panel of Fig. 8). A slope of 1:1 would account for carbonate as calcite. Dolomite, CaMg(CO 3 ) 2 , can also be a source of carbonate ions thus we also include a plot versus the sum of the molar equivalents of Ca 2+ + Mg 2+ . For both the Hawaii and Alaska data, large ion imbalances occur up to ratios of 1:4. When plotted vs. calcium the imbalance is reduced to within +10%. When plotted vs. the sum of the calcium and magnesium the imbalance is reduced to within -11%. An uncertainty of +/−15% can be considered the limit of the filter-based measurement technique or the result of missing ions such as Fe 2+ and H + . We conclude that the missing anion is most likely CO 2− 3 . Some of the tropospheric airmasses analyzed above contained mineral dust as well as high concentrations (> 800 fCi sm −3 ) of the stratospheric airmass tracer 7 Be Jordan et al., 2003b). Stratospheric concentrations of HNO 3 (g) are typically higher than those observed in the FT. As a result, Asian dust lofted beyond the tropopause or into a FT airmass that mixes with stratospheric air could potentially be re-exposed to nitric acid. This would result in additional conversion of HNO 3 (g)→Ca(NO 3 ) 2 in excess of the nitrogen species converted from the emission of anthropogenic precursors (i.e. NO x ). However, when excess sulfate plus nitrate is plotted against 7 Be there is no clear trend between elevated excess NO − 3 and 7 Be (not shown). Thus the higher slope and more robust correlation between nitrate and dust for the subset of upper troposphere-lower stratosphere samples during INTEX-B cannot be attributed directly to secondary exposure to stratospheric nitric acid. Possible explanations include higher initial concentration of anthropogenic NO x mixing with the Asian dust near the source, differing thermodynamic histories for the airmasses (e.g. exposure to high RH) and filter-based sample integration times (5-10 min) longer than that required to analyze fine-scale structure in the tropospheric folds. The results do not preclude the possibility that exposure to stratospheric nitric acid can cause a second set of heterogeneous reactions to take place, but simply that this potential contribution cannot be determined from the INTEX-B measurements.

Heterogeneous chemistry in the nitrate-rich boundary layer of the Central Mexican Plateau
The Mexico City airshed is a complex setting with diverse sources of primary aerosol (soot, organics, dust) as well as secondary aerosol precursors (SO 2 , NO x , VOC's) (DeCarlo et al., 2008;Molina et al., 2008;Salcedo et al., 2006). A detailed analysis of aerosol in the airshed is beyond the scope of this publication. But, motivated by the success of the heterogeneous chemistry analysis over the remote Pacific, we evaluate the nitrogen rich boundary layer of the Central Mexican Plateau (alt <2.0 km AGL), and the FT over the Gulf of Mexico (alt >1.5 AGL), locations where the DC-8 measured the highest concentrations of mineral dust during MILAGRO/INTEX-B. The top row of Fig. 9 are from the DC-8 dataset and indicate a poor correlation between sulfate and calcium. The intercept for the regression can be brought to near zero by subtracting submicrometer sulfate, however, the trend with calcium is still not compelling evidence of heterogeneous reactions. The relationship between NO − 3 and Ca 2+ is highly variable but a general trend with perhaps two branches is evident. Total aerosol chloride from the UNH filters show a relatively robust correlation with the measurements of total calcium. However, in contrast to the samples from the Eastern North Pacific, the chloride from the Mexico City samples is not likely of a marine origin. There is generally sufficient ammonium to partially neutralize the sulfate aerosol (middle-row, Fig. 11). This is consistent with other data collected near Mexico City where the submicron mode was typically neutralized by ammonium DeCarlo et al., 2008;Salcedo et al., 2006). When nitrate and chloride are considered there is insufficient ammonium to neutralize the acid species.
When excess sulfate is plotted versus calcium (bottomrow, Fig. 11) we find large negative values which appear to be associated with relatively fresh fire plumes sampled near Mexico City. The slope and R 2 of the modified regression do not indicate a trend that is sufficiently different from zero but could indicate the presence of gypsum in the Mexico City dusts. After computing excess sulfate + nitrate, there is sufficient ammonium to form both (NH 4 ) 2 SO 4 and NH 4 NO 3 , but insufficient ammonium to consume all the excess acids. The regression against calcium is relatively robust (R 2 = 0.52) and indicates that 22-30% of the calcite alkalinity has been consumed through reactions with nitrogen species. Perhaps the most intriguing result is the improved correlation (R 2 = 0.56) and a further 8% consumption of calcite alkalinity by chloride over Mexico City. The initial dust composition is a relative unknown, though ratios of Na:Ca and Mg:Ca are indicative of crustal material rather than sea salt (not shown). The exceptions stand out as the large  cation-anion imbalances that do not lie along the 1:1 line in the central and right panels of Fig. 10. These data points have geochemical ratios nearer to sea-salt, likely contain primary salts derived from the dry lake beds of the Central Plateau (e.g. Texcoco (Moffet et al., 2008;Moya et al., 2004), and were visually observed as dust sources (C. Mc Naughton) from on board the DC-8 during INTEX-B.

1996-2008 record of aerosol chemistry at the Mauna Loa Observatory
Airborne data collected during a single campaign is effectively a point measurement in both space and time. Here we put the INTEX-B measurements into a broader temporal context by comparing the data to the long-term record from the Mauna Loa Observatory. For the past 13 years the University of Hawaii has conducted measurements of aerosol chemistry at the Mauna Loa Observatory (MLO) . The top row of Fig. 11 plots total aerosol sulfate, nitrate and chloride versus aerosol calcium for the 13-year MLO record. Note that the data are colour coded by scattering Angstrom exponent (red = 3.0; blue = −0.5) and that the y-axis scales differ. We also include a solid and a dashed line which approximately bound the ranges of values for dust concentrations greater than 20 neq sm −3 . The middle row of Fig. 11 plots the degree of neutralization of the acidic aerosol species by ammonium and indicate that aerosol in the FT over the North East Pacific are typically only partially neutralized. Confining our analysis to Ca 2+ concentrations greater than 20 neq sm −3 we plot regressions for excess sulfate, sulfate plus nitrate and excess sulfate and nitrate plus chloride. The more comprehensive MLO data set indicates that 19% of the calcite alkalinity is consumed by sulfuric acid whereas an additional 7% and 2% can be ascribed to reactions with nitric and hydrochloric acid. The 2% increase due to reaction with HCl is actually not significant as the uncertainty in the slope is +/−4%. Regressions including oxalate are similar with   (top-row). Note that the data are shaded by scatteringa angstrom exponent. Scales differ and that the solid and dashed lines are provided as guides to the slope. The middle-row is the degree of neutralization of the acid species by ammonium and are shaded by scattering angstrom exponent. Bottom row compares linear regressions of excess sulfate, suflate plus nitrate, and sulfate, nitrate and chloride versus calcium. Fig. 11. Total aerosol sulfate, nitrate and chloride versus aerosol calcium for the 13-year record of filter based measurements at the Mauna Loa Observatory (top-row). Note that the data are shaded by scatteringa angstrom exponent. Scales differ and that the solid and dashed lines are provided as guides to the slope. The middle-row is the degree of neutralization of the acid species by ammonium and are shaded by scattering angstrom exponent. Bottom row compares linear regressions of excess sulfate, suflate plus nitrate, and sulfate, nitrate and chloride versus calcium. a further 1% consumption of alkalinity but with an uncertainty in the slope of +/−4%. And while variability appears to be driven largely by the association between sulfate and calcium, irreversible sequestration of anthropogenically derived nitrate is significant. Finally, Fig. 11 indicates a few cases where reactions approach the 1:1 alkalinity line. It is these instances where the buffering capacity of the Ca 2+ as CaCO 3 could have been be exceeded and acid processing of the mineral dust is most likely to have occurred (Solomon et al., 2009). Note however that these cases are the exception (∼5 cases) rather than the rule based on the 65 nighttime samples plotted.   , mixed dust and pollution (red) for which there was a 25% reduction in the NMD. A 10% reduction in NMD is simulated in green while the base case of [Boucher and Anderson, 1995] is in magenta and the "water soluble" OPAC aerosol [Hess et al., 1998] is in black. Scattering size distributions at 40% RH (middle) and 80% RH illustrate the significance of NMD and σ g on aerosol optical properties in the accumulation mode. for which there was a 25% reduction in the NMD. A 10% reduction in NMD is simulated in green while the base case of (Boucher and Anderson, 1995) is in magenta and the "water soluble" OPAC aerosol (Hess et al., 1998) is in black. Scattering size distributions at 40% RH (middle) and 80% RH illustrate the significance of NMD and g on aerosol optical properties in the accumulation mode.

The effect of heterogeneous reactions on Asian Dust optical properties
In Fig. 12 we plot the range of mass (volume) distributions for dust (mean +/−1σ , in blue and green) measured over the Eastern North Pacific during INTEX-B, as well as a Saharan distribution (red) measured after long-range transport . The hygroscopic growth factor for Ca(NO 3 ) 2 at 90% RH compared to 40% RH is a factor of ∼1.5. Fully consuming 20 wt% of CaO, MgO and CaCO 3 to produce Ca(NO 3 ) 2 would only increase the volumetric mean diameter of the dust distribution from 2.5 µm to ∼3.0 µm between these two RH's. Computing scattering at 550 nm assuming spherical particles and using Mie theory with a fixed refractive index (1.53-0.0006i (Clarke et al., 2004, Fig. 15b), results in a 7-8% reduction in light scattering due to a reduction in mass scattering efficiency at larger sizes. This small difference in optical properties is negligible compared to the observed variability in the dust distributions and their composition. Additionally, this change in optical properties is small compared to the uncertainty in predicting ambient RH in climate models (Petch, 2001).

The effect of heterogeneous reactions on pollution aerosol optical properties
In the previous section we demonstrate that heterogeneous reactions have a negligible effect on dust optical properties compared to their variability in-situ. However, the presence of supermicrometer dust reduces the mass of secondary aerosol species that condense on primary aerosol of anthropogenic origin; a process that is approximately proportional to aerosol surface area (Howell et al., 2006). This significantly reduces the single scattering albedo of the accumulation mode aerosol  as well as f(RH) (Howell et al., 2006). Here we compare the reduction in ambient light scattering due to a reduction in accumulation mode NMD, to the effect of enriching the accumulation mode in relatively hydrophobic organic species, i.e. a reduction in non-refractory (OM, sulfates and nitrates) accumulation mode mass due to the presence of dust.
The left panel of Fig. 13 shows five number distributions examples. The blue and the red lines are the observations from INTEX-B for the pollution dominated aerosol and for the mixed aerosol types. The green distribution is a simulated distribution with only a 10% reduction in the pollution NMD. The magenta line is the "base case" from the sensitivity analysis by (Boucher and Anderson, 1995) while the black line is the "water soluble" aerosol type from the popular Optical Properties of Aerosols and Clouds (OPAC) parameterization (Hess et al., 1998). The log-normal fitting parameters used to describe these distributions are found in Table 5. Scattering, calculated at 40% and 80% relative humidity, is shown in the middle and right hand plots of Fig. 13. Table 6 compares light scattering for the blue, green and red distributions by evaluating their integrals at 10%, 80% and 95% RH. From these calculations it is clear that small reductions (10-25%) in the distributions NMD has a large influence on aerosol optical properties, reducing scattering by as much as 30-70%. This concern is particularly relevant to global climate models that use static aerosol size distributions (e.g. OPAC), or those that use modal or sectional approaches that cannot capture the dynamic nature of the aerosol number distribution.
In Fig. 13 and Table 6 we modeled the optical properties by assigning a growth factor consistent with an ammonium sulfate type composition and by varying the refractive index as a function of RH. The parameter γ , used in the f(RH) function (Eq. 1), implicitly incorporates the effects that water uptake, aerosol size, and changing refractive index have on aerosol optical properties. For the pollution distribution with an NMD 0.10 µm (blue in Fig. 13), the ammonium sulfate type composition results in a γ value of ∼0.76. Strictly speaking it is not appropriate to apply this γ to the other two distributions (NMD = 0.090 µm and 0.075 µm) as their γ values are uniquely dependent on their size distributions and composition. However, Fig. 14 plots the explicitly calculated scattering values as a function of RH compared to those generated using a static γ of 0.76 for each distribution. Table 7 tabulates the difference between the static γ values compared to the explicitly calculated scattering values at 10%, 80% and 95% RH. As indicated, the static γ method tends to over predict scattering by ∼10-25% at relative humidities below 95%. The static γ method does however have an advantage. Since actual aerosol composition, their diameter growth factors -g(RH), and refractive indices are poorly constrained, we can use the static γ approximation to evaluate the effects of size versus composition on aerosol optical properties for these examples.
The INTEX-B, free troposphere, total aerosol, average f(RH) γ is 0.49-0.52 (Table 4). This value is effectively identical to the ACE-Asia average 0.51 for fine mode dominated total aerosol (Anderson et al., 2003), indistinguishable from pollution dominated FT values over North America (0.47+/−0.7) (C. Mc Naughton, unpublished data, INTEX-NA), but lower than the ACE-Asia value, 0.63¸in the pollution dominated MBL (Carrico et al., 2003). The ACE-Asia Table 8. Changes to light scattering at 80% relative humidity due to a reduction in aerosol number median diameter (blue). These reductions are compared to a reduction in accumulation mode aerosol hygroscopicity from γ = 0.5 to γ = 0.48 (red).  Figure 14 -Comparison of scattering as a function of RH calculated using Mie theory and assuming an (NH 4 ) 2 SO 4 composition and an RH dependent refractive index (solid lines) for the three distribution in Figure 13 compared to those using an estimated γ of 0.76 (dashed lines).
allows us to estimate the ratio of particulate organic matter to POM + SO 2− 4 according to: where C o is the mass of POM and C s is the mass of sulfates. POM can be further divided into both semivolatile (SVolOM) and refractory (RefOM) organic components . For an f(RH) γ of 0.50, F o is 0.57. As discussed in section 4.1 the INTEX-B average refractory volume fraction is ∼12-16%, +/−7%. Assigning a density of 1.3 g cm −3 to both the refractory and semivolatile OM mass material    Table 8 briefly summarizes the comparison between reductions in light scattering due to aerosol size compared to changes in composition. A 10-25% reduction in the size distribution NMD leads to a 30-70% reduction in light scattering. After reducing the accumulation mode semi-volatile inorganic (i.e. sulfates) and organic mass by 50%, light scattering at 80% RH only decreases by ∼3%.
Thus we conclude that reductions in ambient light scattering, due to competition between pollution and dust aerosol surface area for condensing secondary species, is dominated by reductions in median aerosol size, and not changes to accumulation mode aerosol composition.

The effect of heterogenous reactions on NO y
In Sect. 4 we established that the heterogeneous reactions are occurring between gas-phase nitrogen species (e.g. HNO 3 , NO 2 ) and mineral dust during long-range transport from Asia. These reactions result in the sequestering of gas-phase compounds in the supermicrometer particulate phase. These relations are replotted in Fig. 15 to examine the effect of Asian dust on NO y . Here NO y is defined as the sum of the available in-situ measurements of: NO, NO 2 , HNO 3 , PNs and ANs. The left panel shows that nitric acid is converted into aerosol nitrate and is reduced to ∼20% of the sum of these two species at high dust concentrations. Pollution dominated airmasses are rich in NO y but particulate NO − 3 is <∼20% of the total nitrogen species (top-center and right of Fig. 15). As dust mass increases, particulate nitrate is created from NO y and becomes a larger fraction (20-40%) of total nitrogen. Recall that these estimates are somewhat conservative as the DC-8 measurements of gas-phase HNO 3 are likely contaminated with aerosol NO − 3 (Miyazaki et al., 2005;Zondlo et al., 2003).
Whereas the filter-based measurements of aerosol chemistry are slow (300-600 s), the HiGEAR size distribution measurements are relatively fast (3 seconds). In Fig. 5 we established a robust (R 2 = 0.90) regression of 11 wt% for the CaCO 3 present in the Asian dust samples. In Fig. 7, 9 and 11 we show that ∼5-30% of the CaCO 3 has been consumed by reacting to form Ca(NO 3 ) 2 . Assuming an 11wt% for the dust calcite content and 25% consumption of this alkalinity, we explore the ratios of aerosol nitrate to HNO 3 and NO y using 60-second averages of HNO 3 , NO y and supermicrometer dust. Data are restricted to the Pacific free troposphere (alt. >2.0 km, O 3 <200 ppbv), west of 125 • W where fast (90second) submicrometer sulfate (UNH mist chamber) concentrations are above 1.0 µg sm −3 . The results of this simulation are shown in the bottom panels of Fig. 15 using red circles. In general we have replicated the shape of the measured relation. The main discrepancy is that the simulated ratios have values higher than those observed using the slower chemistry measurements. This can be attributed to the fact that the tail of the dust distribution has been shown to extend below the 1-µm cut size used to isolate the supermicrometer dust and nitrate mass (see also Clarke et al., 2004). Also note that the use of 25% conversion of the CaCO 3 is nearly an upper bound and that some nitrate is present as submicrometer NH 4 NO 3 and has not been accounted for.
We can also examine these relationships using the fast chemistry measurements from the NSF/NCAR C-130 collected off the US west coast during INTEX-B. The nitrogen ratios are shown in the upper three panels of Fig. 16 for measurements aboard the C-130. These data are stratified using the same criteria as Dunlea et al. (2009), but do not show the same characteristics as the INTEX-B data (e.g. reduction in NO y :total nitrogen versus Ca 2+ or dust). The reason for this difference is that, although fast, the HR-ToF-AMS only measures non-refractory NO − 3 in the submicrometer mode, whereas most (∼>70%) of the nitrate is dust-bound in the supermicrometer mode (Kline et al., 2004). Still, the far right panel of Fig. 16 shows an interesting trend in the AMS measurements of NO − 3 . There is a slight reduction of NO y :total nitrogen as a function of dust mass. This could be attributed to Ca(NO 3 ) 2 detected by the HR-ToF-AMS below 1.0-µm, in the lower-tail of the dust distribution. As in Fig. 15, we can estimate coarse-mode nitrate using the fast size distribution measurements aboard the C-130. This is shown in the bottom three panels of Fig. 16 and is an approximation to the chemistry-based measurements presented in the top row of Fig. 15.
The upper panel from Fig. 15 and the lower panel from Fig. 16 give us the most complete picture of the heterogeneous reactions we have observed. The observations support the findings of Song and Carmichael (2001) and Bian and Zender (2003), both of whom estimate that heterogeneous reactions in the presence of dust reduces HNO 3 by >50%.
Second, the presence of dust irreversibly sequesters NO y species in the particulate phase. Without dust (<5 µg sm −3 ) NO y accounts for ∼>85% of the total nitrogen. When dust concentrations exceed ∼10 µg sm −3 , supermicrometer nitrate formation reduces the NO y fraction such that it accounts for 60-80% of total nitrogen. This is in good agreement with Phadnis and Carmichael (2000) who predicted that "the presence of dust (increases) particle nitrate levels in East Asia by ∼40%".

Conclusions
In spring 2006 the NASA DC-8 and the NSF/NCAR C-130 measured trace-gases and aerosol over Mexico and the Eastern North Pacific as part of MILAGRO and INTEX-B. Airmasses transported long-range from East Asia were sampled in the free troposphere and found to contain Asian pollution and Asian pollution mixed with dust. Accumulation mode and supermicrometer size distributions are summarized and fit using log-normal functions. The presence of dust is found to reduce the median diameter of the accumulation mode aerosol by up to 25% due to competition for condensing secondary aerosol species. Adding secondary aerosol mass to the supermicrometer mode through condensation and/or heterogeneous reactions has a negligible effect on dust hygroscopicity and optical properties. However, altering the median diameter of the accumulation mode aerosol by 10-25% can reduce light scattering values by factors of 30-70% at 80% relative humidity. Accounting for the relative enrichment of hydrophobic species in dust-influenced accumulation mode aerosol results in a further reduction in light scattering of ∼3% at 80% relative humidity. Thus we conclude that the reduction in accumulation mode aerosol size, and not subtle changes to its composition, dominates the effect of mineral dust on accumulation mode aerosol optical properties. In order to better predict the influence of heterogeneous reactions on aerosol direct and indirect effects, climate modelers are encouraged to improve the dynamic representation of the aerosol size distribution and include source dependent chemical composition.
Recent modeling studies have been found to make incorrect assumptions regarding the solubility of Asian dust. Literature review, and in-situ measurements presented here, supports a CaCO 3 wt% of between 3.6 and 21% with an average value near 12% for natural Asian dust aerosol. The remaining mineral dust is composed of aluminosilicates and metal-oxides whose water uptake and activation in cloud is a strong function of size, aggregate composition, and exposure to reactive sulfur and nitrogen compounds (Grassian, 2001;Kelly et al., 2007;Matsuki et al., 2009). TRACE-P (Jordan et al., 2003a], ACE-Asia (Kline et al., 2004), MILA-GRO and INTEX-B results, as well as long-term measurements at Mauna Loa Observatory show that only 5-30% of the CaCO 3 present in Asian dust is consumed to form more soluble Ca(NO 3 ) 2 . As a result, only 1.2-3.6% of the dust mass could be considered "soluble" due to heterogeneous reactions. A further consumption of ∼2-4% of the calcite alkalinity can be attributed to reactions with HCl over the Pacific with as much as 8% occurring in the boundary layer near Mexico City.
Reacting NO 2 and/or HNO 3 with CaCO 3 results in an irreversible sequestering of gas-phase nitrogen species into the particulate phase. In the presence of dust, nitric acid concentrations are reduced to <50% of total nitrate (nitric acid plus particulate nitrate). NO y as a fraction of total nitrogen (NOy plus particulate nitrate), is reduced from >85% to 60-80% in the presence of dust. These observations may have a number of implications for atmospheric oxidants which deserve further study.