Rapid Evolution of Aerosol Particles and their Optical Properties Downwind of Wildfires in the Western U.S.

Abstract. During the first phase of the Biomass Burn Operational Project (BBOP) field campaign, conducted in the Pacific Northwest, the DOE G-1 aircraft was used to follow the time evolution of wildfire smoke from near the point of emission to locations 2–3.5 hours downwind. In nine flights we made repeated transects of wildfire plumes at varying downwind distances and could thereby follow the plume's time evolution. On average there was little change in dilution-normalized aerosol mass concentration as a function of downwind distance. This consistency hides a dynamic system in which primary aerosol particles are evaporating and secondary ones condensing. Organic aerosol is oxidized as a result. On all transect more than 90 % of aerosol is organic. In freshly emitted smoke aerosol, NH4+ is approximately equivalent to NO3−. After two hours of daytime aging, NH4+ increased and is approximately equivalent to the sum of Cl−, SO42− and NO3−. Particle size increased with downwind distance causing particles to be more efficient scatters. Averaged over nine flights, mass scattering efficiency increased in ~ two hours by 56 % and in one fight doubled. Coagulation and material transport from small to large particles are discussed as mechanisms for increasing particle size. As absorption remained nearly constant with age the time evolution of single scatter albedo was controlled by age-dependent scattering. Near-fire aerosol had a single scatter albedo (SSA) of 0.8–0.9. After one to two hours of aging SSAs were typically 0.9 and greater. Assuming global-average surface and atmospheric conditions, the observed age-dependence in SSA would change the direct radiative effect of a wildfire plume from near zero near the fire to a cooling effect downwind.



Introduction
Aerosols from wildfires alter Earths radiation balance by their direct interaction with sunlight and by indirect effects mediated through perturbations on clouds and precipitation. In the direct effect, heating is the result of light absorption by black carbon (BC), brown carbon (BrC) -including tar balls, and dust. 25 Cooling occurs when aerosol particles scatter light upwards so that sunlight that would otherwise be absorbed by Earth or within the atmosphere, escapes to space. The transition between heating and cooling depends on intrinsic aerosol properties such as aerosol single scatter albedo (SSA= light scattering/light extinction) and extrinsic factors, including surface albedo, solar zenith angle, and cloud cover (Chylek and Wong, 1995;Nemesure and Schwartz, 1995;McComiskey et al., 2008). 30 On a global basis, wildfire aerosols are estimated to have a net direct radiative effect of 0.17 watts m -2 , 1-σ uncertainty = -0.45 -+0.15 W m -2 (Bond et al., 2013) The rapid evolution of aerosol optical properties of fresh smoke contributes to the radiative effect uncertainty (Yokelson et al., 2009;Akagi et al., 2012;May et al., 2015;Vakkari et al., 2018;Selimovic et al., 2019). In this study we are concerned with 35 changes in wildfire optical properties that occur in the near-field, extending from the time it takes a smoke plume to rise to aircraft-sampling altitude to circa three hours downwind. These first hours are a dynamic period as the plume's initial store of reactive VOCs and NOx are largely intact and can generate secondary https://doi.org/10.5194/acp-2020-239 Preprint. Discussion started: 20 April 2020 c Author(s) 2020. CC BY 4.0 License. range over which light scattering per unit mass of aerosol (MSE = mass scattering efficiency) varies several-fold. 75 The first phase of the Biomass Burning Operational Project (BBOP; Kleinman and Sedlacek, 2016) field campaign was conducted in the temperate Pacific Northwest. Twenty-one research flights were conducted between July and September 2013. Primary targets were i) wildfires in which the time evolution of aerosols and trace gasses could be determined from measurements at multiple distances between the fire 80 and locations several hours (smoke physical age) downwind and ii) overflights of a ground site at Mount Bachelor Observatory (MBO) in Oregon and areas upwind at times when MBO was impacted by wildfire smoke (Collier et al., 2016;Zhou et al., 2017;Zhang et al., 2018). One MBO flight coincided with a SEAC 4 RS flight (Liu et al., 2017). In the second phase of BBOP, during October 2013, agricultural burns in the lower Mississippi Valley were sampled. In both phases of BBOP, TEM samples were analyzed for 85 tar balls (TBs) as reported by Sedlacek et al. (2018a) and Adachi et al (2018;. In this study we use data from five wildfires that were collected during nine flights. Pseudo-Lagrangian sampling allows us to determine the rate of change of aerosol, gaseous, and optical quantities as a function of transport time or photochemical age. Extensive variables are normalized to CO to account for 90 dilution. We use the negative logarithm of the ratio of NOx to NOy (-Log10 (NOx/NOy)) as a metric for photochemical processing. Of primary interest are time dependent changes in the concentration of submicrometer diameter aerosol, mass scattering efficiency (MSE), and mass absorption coefficient (MAC).
The latter two quantities yield a local determination of SSA. Though there was only a small effect of aging upon total aerosol mass concentration, within a couple of hours there were i) changes in the concentrations 95 of inorganic species and in organic speciation as judged by atomic ratios and mass peaks at m/z = 43, 44, and 60 and ii) a significant increase in scattering and MSE causing plumes to transition from a near-neutral radiative forcing to one that is cooling. Because observed size distributions were limited to Dp < 210 or 260 nm and most scattering is from larger particles we cannot (without extrapolating the size data) calculate MSE for comparison with the observed ratio of scattering to aerosol mass. Instead, we have 100 determined the diameter within the accumulation mode where the number size distribution has a peak and using that diameter as a metric for particle size we show that MSE is positively correlated with particle size and negatively correlated with backscatter ratio (total scattering/scattering into rear hemisphere) and scattering Angstrom exponent. These correlations between MSE and particle size and between MSE and scattering properties give support to the hypothesis that MSE increases with age because coagulation and/or 105 chemical and microphysical processes shift particle mass from small diameters to the large diameters where scattering is more efficient. https://doi.org/10.5194/acp-2020-239 Preprint. Discussion started: 20 April 2020 c Author(s) 2020. CC BY 4.0 License.

Experimental methods
In the first phase of BBOP the DOE G-1 aircraft sampled wildfire plumes in the temperate forests 110 and range lands of the western U.S., focusing on time evolution from near a fire to a few hours transport time downwind. An enhanced suite of instruments at Mount Bachelor Observatory (MBO), Oregon provided a continuous record of BB plume properties. In coordinated flights the G-1 sampled plumes that later impacted MBO (Collier et al., 2016;Zhou et al., 2017) and one flight in Southern Oregon coordinated with the SEAC 4 RS DC8 (Liu et al., 2017). 115 An extended deployment, June 7, 2013 -Sept. 13, 2013, was made possible by basing the G-1 at the home of the ARM Aerial Facility (Schmid et al., 2014) in Pasco, WA. The choice of Pasco as a center of operations was justified by examining alternate sites, covering most of the U.S. Western forest fire region. For each location the frequency of fires was determined based on daily emissions of CO2, CO, 120 NMHC, and PM2.5 from MODIS fire products and the FINNv1 (FIre INventory from NCAR) emission inventory . Annual variability was estimated from 10 years of data. Fire counts and emissions were summed over the G-1s 500 km operational range from the candidate sites.

Instruments 125
Schmid et al (2014) review the G-1 aircraft and the instrument suite provided and maintained by the ARM Aircraft Facility. The following discussion and the list in Table 1 is restricted to instruments used in the present study. The G-1 instruments used in BBOP are also summarized by Liu et al. (2016).

SP-AMS 130
The BBOP field campaign was among the first aerial deployments of the Soot Particle Aerosol Mass Spectrometer (SP-AMS). For an overview of the SP-AMS, see Onasch et al (2012). In brief, the SP-AMS is a high-resolution time-of-flight aerosol mass spectrometer (HR-tof-AMS) to which a high-power CW YAG laser is added for vaporizing light absorbing refractory aerosol such as BC. The SP-AMS can be operated in laser-off or laser-on modes. With the YAG laser off, non-refractory (nr) particles are vaporized 135 by impacting a heated target, typically at 600 ºC; performance characteristics such as collection efficiency (CE) and fragmentation patterns are identical to an HR-tof-AMS. In laser-on mode, strongly absorbing aerosol components such as BC are vaporized before encountering the heated target. Non-absorbing species that coat BC particles are vaporized by heat transfer as happens in an SP2 (Schwarz et al., 2010). Brown carbon (BrC), so-called because of its short-wavelength absorption, doesn't absorb the 1064 nm YAG laser 140 unless there is a long wavelength absorption tail. The thermal vaporizer was left on when operating in laser-on mode, thereby allowing nr-aerosol to be quantified, but at the expense of not being able to selectively detect the coating on BC particles.
The SP-AMS and constant pressure inlet were operated as described by Collier et al., 2016. One Hz 145 data was acquired in "Fast-MS" V-mode. Instrument-backgrounds were determined by alternating between https://doi.org/10.5194/acp-2020-239 Preprint. Discussion started: 20 April 2020 c Author(s) 2020. CC BY 4.0 License. 52 seconds of signal collection and 8 seconds with the aerosol beam blocked. Most SP-AMS data was acquired in laser-on mode. In order to investigate sampling strategies, we sometimes alternated between laser-on and laser-off operation, either minute by minute or for repeated plume crossings. The collection efficiency of the SP-AMS during laser-off operation was determined to be 0.5 via comparisons between 150 aerosol mass concentrations measured on the G-1 with similar measurements at MBO during overflights (see Fig. S2 of Collier et al., 2016). In each case we normalized non-refractory aerosol concentration to CO to account for changes in plume dilution between transects. The average ratio for 16 transects; laser-on to laser-off was 1.52. This procedure yielded a laser-on CE of 0.76 with a standard deviation of 0.07.

155
We assume, as done implicitly in other studies, that TBs contribute towards total non-refractory organics with the same detection efficiency as other organic aerosol. This assumption needs to be tested for ambient aerosol of the types observed in BBOP, especially since TBs were observed to have a mass fraction of 25% -40% of PM1 in aged smoke (Sedlacek et al., 2018a). As a first step, the thermal stability of TBs was investigated by electron microscopy of BBOP samples affixed to a heating stage in which the 160 temperature was ramped up over 10 minutes to the 600 °C temperature of the SP-AMS thermal vaporizer . At 600 °C, 30% of the TB mass collected on TEM grids was not vaporized. If the results of these slow-heating experiments are applicable to the thermal flash vaporization occurring in an AMS, a fraction of nominal nr-organic mass is not vaporized and therefore not detected. We estimate this fraction as being of order 10% (25-40% mass fraction TB  30% not evaporated in TEM experiments). 165 Evidence to the contrary, namely, that TBs are detected with the same efficiency as other organic aerosol by an AMS is that TBs have the correct magnitude and volatility (up to 200 °C) to coincide with BBOA-3, a low volatility PMF factor of BB smoke sampled at MBO by Zhou et al. (2017). Sedlacek et al. (2018b) andCorbin andGysel-Beer (2019) showed that, depending on laser power, 170 TBs can be charred by an SP2 YAG laser identical to that used in the SP-AMS. The charred residue has the potential to be detected as BC in a SP-AMS. BC mass concentrations in aged, temperate wildfire plumes were observed to be an order of magnitude less than TBs (Sedlacek et al., 2018a) and therefore BC concentrations are susceptible to large errors if even a small fraction of TB mass is mis-identified as BC. In this study we use BC concentrations as observed by an SP2. In keeping with the nomenclature of the SP2 175 community, the light absorbing, incandescing species measured by the SP2 will be referred to as rBC.
In recent measurements at the USDA Fire Sciences Laboratory, Lim et al (2019) used an AMS to analyze BB generated aerosols soon after emission and then again after exposure to oxidants in a chamber simulating up to several days of photochemical aging. The AMS collection efficiency (CE) was 180 determined by comparison with SMPS measurements of aerosol size distribution, integrated to give a total volume. The AMS CE was 0.54 for fresh emissions, decreasing to 0.40 for multi-day aged aerosol. This decrease correlated with a decrease in aerosol volatility as measured by a thermal denuder. Aged particles https://doi.org/10.5194/acp-2020-239 Preprint. Discussion started: 20 April 2020 c Author(s) 2020. CC BY 4.0 License.
were interpreted as being more viscous and more likely to bounce off the AMS oven and thereby escape detection. If we assume the change in CE found by Lim et al (2019) our assumption of a constant CE 185 results is a 35% underestimate for the change in aerosol mass concentrations in aged air relative to fresh emissions.
Plume observations in BBOP covered a maximum aging time of 3.5 hours. During aging we find, as have others (e.g., DeCarlo et al., 2010;Heald et al., 2010;Ng et al., 2011) an increase in the aerosol O to 190 C ratio (O:C); a compositional change that has been observed to be correlated with increased viscosity.
This change in viscosity in aged smoke is consistent with the conclusion of Sedlacek et al (2018a) that high-viscosity tar balls, found only after aging, are processed primary particles. The dependence of CE on O:C has been determined for laboratory biomass burns by Onasch (personal communication). Combined with the O:C changes found in BBOP (presented below) we obtain an estimate that aged BBOP plumes can 195 contain up to 15% less aerosol mass than measured based on a fixed CE. We have not incorporated an agedependent CE into our calculations. It is likely to be small for the range of aging that we cover and to the best of our knowledge AMS observations of ambient biomass burn plumes have not shown systematic changes in CE.

SP2
A single particle soot photometer (SP2; Revision D), Droplet Measurement Technology) was used to measure the number concentration and size distribution of refractory particles. Aerosol particles are subjected to high intensity CW 1064 nm light from a YAG laser, heating particles that absorb at this wavelength to the point of incandescence. Color temperature of the incandescence is used to discriminate 205 between BC and other absorbers such as mineral dust. The size distribution of rBC particles with volume equivalent diameters between ~ 80 and 500 nm was determined based on a rBC density of 1.8 g cm -3 .

FIMS 210
Aerosol size data used in this study are from the FIMS (Fast Integrated Mobility Spectrometer) (Olfert et al., 2008;Wang et al., 2017), designed to provide information similar to an SMPS. Because the FIMS measures particles of different sizes simultaneously instead of sequentially as in a traditional SMPS, it provides aerosol size spectra with high time resolution (i.e. a size spectrum is acquired in 1 sec vs. 60 secs for the SMPS). The minimum size particle classified had a diameter of 20 nm. The classifying 215 voltage in the FIMS is reduced at high altitude to prevent arcing. At the altitudes where plumes were sampled (typically 2500 -3000 m above msl) the upper size limit of the FIMS deployed during BBOP was limited to 210 nm or 260 nm as a result of the reduced classifying voltage. The FIMS size range usually encompassed the peak of the particle number size distribution, dN/dLogDp. FIMS data near the peak in dN/dLogDp (restricted to the accumulation mode, between 100 nm and 260 nm) were used to estimate the diameter (between bin boundaries) at which the peak occurs. We use this derived diameter as a surrogate for the geometric mean diameter, DGEO, of the accumulation mode.

Scattering
Aerosol scattering, and back scatter ratio were determined from a three wavelength (450, 550, 700 225 nm) TSI 3563 nephelometer. Unless otherwise noted, scattering will refer to measurements at 550 nm.
Flow rate and internal volume limited response time to 2 sec. In the G-1 data set, corrections for the deviation of the instrument from a cosine response and the smaller correction for not detecting photons scattered into a 7 degree forward and 10 degree backward cone were applied using the sub μm formulas from Anderson and Ogren (1998). Size spectra up to 50 μm were determined from a Cloud, Aerosol, and 230 Precipitation Spectrometer (CAPS) probe mounted outside the fuselage. The default CAPS diameter bins for liquid water were mapped onto bins corresponding to a refractive index of 1.55, as appropriate for an organic dominated aerosol. According to Mie calculations, supra-micrometer particles (1 μm < Dp < 5 μm) contributed only a couple of percent to total scattering. Nephelometer measurements were therefore interpreted as being due to sub μm particles. 235

Light Absorption
Aerosol light absorption at 532 nm was measured with a Photothermal Interferometer (PTI; Sedlacek and Lee, 2007). This instrument relies on an interferometer to detect a change in optical path length caused by aerosol light absorption and subsequent degradation of absorbed energy into heat. As such 240 the PTI is not susceptible to filter based artifacts affecting the PSAP and similar devices (Lack et al., 2008).
For the calculation of single scatter albedo (SSA) we assume that absorption varies as 1/λ, as is approximately correct for BC. PTI measurements are thereby decreased by 3% in order to adjust absorption measurements to 550 nm, the wavelength at which scattering data is available. Based on the 3-PSAP, it is likely that there is some BrC absorption at this wavelength but in view of the smallness of the 245 adjustment it is neglected.

Trace Gases
NO, NO2, and NOy were measured in a 3-channel chemiluminescence instrument (Air Quality Design, Golden, CO). Ambient NO was measured in the first channel with a small delay volume to give 250 simultaneity with the NO2 channel. NO2 conversion occurs in a low-pressure cell with LED irradiation at 390 nm. Conversion efficiency for NO2 was measured to be 0.50 +/-0.03. NOy (defined as odd-nitrogen species including aerosol nitrate) was measured via a Mo converter heated to 350 o C located externally on a pylon affixed to a window blank. Conversion efficiency for NO2 was 0.98 +/-0.02 and earlier tests have shown similar efficiencies for HNO3 and organic nitrates (Williams et al., 1998).

Coincidence, Flow Control, and Dilution system 265
It was anticipated that aerosol number concentrations close to wildfires would exceed the coincidence thresholds of several particle-resolved instruments (Table 1). At threshold a specified small percentage (e.g. 10%) of aerosol particles cannot be distinguished one from the next, resulting in missed counts or particles sorted into incorrect size bins. To mitigate coincidence errors, zero-air from a cylinder was used to dilute ambient air in a ratio that was as high as 10 to 1. Dilution rates were changed in-flight to 270 maintain a high instrument sensitivity in clean air. CPC data were corrected for coincidence.

Data
Concentrations and optical measurements are reported at STP, 0 °C and 1 atm. pressure. Particles sizes are diameters. For each instrument and each flight, one Hz data were time shifted to maximize the 275 correlation with light scattering, a time standard selected because of its availability on all flights and because it exhibits high correlations with most other quantities. Time shifts account for instruments sampling the same air parcel at different times. At a sampling speed of 100 m s -1 , km-scale plumes appear in the data record as rapidly changing signals. Time-shifts of 1-2 seconds are readily apparent as a degradation in correlation when comparing instruments. Maximizing correlations, however, does not 280 accurately compensate for varying response times. In so far as possible, we rely on average values across a plume, which are relatively insensitive to time response.
Except as noted, aerosol concentrations are the mass of non-refractory components as measured by the SP-AMS and are referred to as nr-PM1. Units are μg/m 3 at STP. By presenting nr-PM1 we do not limit 285 the aerosol concentration measurements by the requirement that the SP2 was also acquiring data at the same time. Mass differences between non-refractory and total aerosol are ~ 0.5 -2.5%. References to aerosol number concentration (particles/cm 3 at STP) will be explicit to distinguish from aerosol mass.
Flights are identified by month (m), day (dd), and an "a" or "b" for the 1 st or 2 nd flight of a day, e.g., 290 821b was the second flight on Aug. 21, 2013. All times are UTC. Local time was Pacific-daylight savings and is given by UTC -08:00. Local noon over our sampling region on 1 August ranged from 19:43 UTC in the east to 20:15 UTC in the west. Data from the BBOP field campaign have been archived at (https://www.arm.gov /research/campaigns/aaf2013bbop. Wildfire flights analyzed for aging in this study had to meet three criteria. 1) Sampling included 305 fresh emissions, estimated as having an atmospheric residence time less than 30 minutes. That emissions were fresh was determined by the observation of a high concentration, compact plume with a high NOx to NOy ratio. 2) Measurements were made at downwind distances where significant aging was expected.

Flights and Wildfires
Transport between regions with fresh and aged pollutants had to be consistent with in-plume wind measurements. 3) Measurements of aerosol mass, light scattering, and CO were required. 310 Nine pseudo-Lagrangian flights sampled smoke from five wildfires (Tables 3 and 4) for which the time evolution of smoke could be followed from near its source to locations several hours downwind.
Evidence that fire characteristics were reasonably steadyan implicit assumption upon which Lagrangian analysis depends -comes from flights in which two sets of transects were repeated with about a one-hour 315 delay between samples at nearly the same location. True Lagrangian sampling was not possible. Leaving aside the difficulty of identifying and following a particular air parcel, the time over which this could be done is restricted by the G-1's limited sampling time of one to two hours on station. Ground tracks for flights 730b and 821b are shown in Figs. 3 and 7, respectively. The other seven flights satisfying our pseudo-Lagrangian criteria are shown in Figs. S1 -S7. For the nine-flight ensemble, plumes were sampled 320 at downwind distances between a few km and 90 km, at which point up to 3.5 hours of aging occurred.
The minimum downwind distance is dictated by the distance needed for the plume to reach aircraft altitude or, in some cases, by flight restrictions. Seven of the nine flights had NOx and NOy measurements that were used to calculate photochemical age.

325
Several flights resemble regular grids containing two sets of up to 6 cross-plume segments. The spacings between segments represents 20 -60 minutes of plume aging. Other flights are more free-form and in one case (821b) there is an extended along-plume segment that has been split into pieces with different ages. The ground track figures indicate the portions of each flight, 106 in total, over which averages are taken. We will refer to these cross-plume and along-plume flight segments as "transects". 330 Transects were defined to be in the wildfire plume at a relatively narrow range of distances from the parent wildfire. Vertical plume structure was not explored because to do so would have been at the expense of horizontal spatial coverage. Wildfire sampling was mostly at constant altitude, at least 1000m above terrain, and insofar as possible near the altitude with maximum concentrations. Unlike small prescribed burns, or observations from a far-downwind vantage point, the wildfires sampled in BBOP have a 335 significant spatial extent in comparison to our measurement domain. There were often hot spots (secondary fires) in nominally downwind locations. Thus, scatter in our measurements as a function of downwind aging is expected.
Data for the nine pseudo-Lagrangian flights comes from the five wildfires listed in Table 3. 340 Descriptions of these fires are available on National Fire Service Incident reports (https://fam.nwcg.gov/fam-web/hist_209/hist_2013_r_209_gacc_sprd?v_gaid=NW. Emission factors for the Colokum Tarp flights (730a and 730b) are compared with SEAC 4 RS and other data sets by Liu et al (2016).

Time Evolution
The time evolution of fire emissions is calculated from plume measurements at varying downwind distances. We use transect-averaged quantities from which backgrounds have been subtracted. In the example of species X, an excess concentration is 350 Xi is an average of X over transect i and XB an average over a relatively unpolluted background region near the smoke plume. If Xi is an intensive variable, such as an AMS mass ratio (f43, f44, f60, O to C ratio, or 355 H to C ratio), a scattering Angstrom exponent, or a backscatter ratio, transect-averages are constructed using nr-PM1 as a weighting function. Problems and alternate ways of determining background are discussed by (Yokelson et al., 2013;Briggs et al., 2016;Garofalo et al., 2019). To account for plume dilution or conversely sampling a more concentrated plume region downwind, we divide measured concentrations by a conservative tracer, CO. A normalized excess concentration is given by 360 A ratio of Xi to another quantity, Yi, is calculated as In any ratio, such as in Eqs. 2 and 3, a common set of data points are used for numerator and denominator.
Except as noted, ratios involving aerosol mass are based on non-refractory aerosol measured with the SP-AMS, i.e. nr-PM1. Equations (1) to (3) apply to variables that are expressed as concentrations, mixing 370 ratios, or inverse lengths. The meaning should be clear from the units used.
Effects of aging are determined from changes in normalized excess ratios as a function of photochemical age or transport time. We define photochemical age by the conversion of NOx to oxidation products, expressed as -Log10 (ΔNOx/ΔNOy) (Olszyna et al., 1994;Kleinman et al., 2008;DeCarlo et al., 375 2010). In the case that NOy is conserved and NOx is lost primarily by OH + NO2 → HNO3, photochemical age, so defined, would yield [OH]*time; and given a trajectory-based time, would yield an average OH concentration. We, however, refrain from inferring an OH concentration from NOx and NOy measurements.
We observe an apparent loss of NOy in fresh plumes (e.g., Neuman et al., 2004). Oxidation of NOx is more rapid than expected from OH+NO2 (Mebust et al., 2011) which can be due to the known high yields of 380 PAN (Alvarado et al., 2010;Briggs et al., 2016;Liu et al., 2016). As will be shown, -Log10 (ΔNOx/ΔNOy) is a useful metric for chemical processing as it varies in the same sense as age-related changes in organic aerosol speciation.

Results 385
Brief summaries of the 9 pseudo-Lagrangian flights are given in Table 4. Transect average modified combustion efficiency (MCE) observations for the 6 pseudo-Lagrangian flights with CO2 observations are tightly grouped between 0.86 and 0.92, close to MCE=0.9, traditionally taken as a transition point between mostly burning and mostly smoldering fires . The relation between MCE and aerosol composition for two regional BBOP flights has been discussed in conjunction with measurements at MBO 390 by Collier et al 2016. Amongst the 9 flights, emission intensity as determined by peak values of nr-PM1, CO, and light scattering varied by about an order of magnitude. Three fires were sampled on multiple flights, the Mile Marker 28 fire (725a and 726a), Colockum Tarps fire (730a, 730b, and 809a) and the Pony Fire Complex (813a and 814a). Plumes from the two Pony Fire Complex flights tended to resemble each other and have a similar age dependence (see figures, below), more so than the Mile Marker 28 and 395 Colockum Tarps flights. In the most aged transects, background was an appreciable fraction of plume values for nr-PM1, CO, and scattering. For several flights background subtraction was problematic for CO2 and inorganic aerosol components.

Plume Age, Concentration, and Dilution 400
Figure 2 shows a comparison between NOx/NOy based photochemical age and atmospheric transport times. Photochemical age increases with downwind distance, but these two metrics of atmospheric processing are not directly proportional, nor should they be. In older plumes, NOx is depleted, and photochemical age tends to level off whilst distance is not bounded. Close to the fire transects have an age ranging from 0.1 to above 0.5. The higher age values are generally from smaller, perhaps more rapidly 405 evolving, fires located near the main source of smoke. In the absence of other information, these transects are assigned a downwind distance equal to that of nearby less aged transects. NOy data is missing from the 809a and 813a flights. Photochemical ages for transects on these flights have been generated from downwind travel times and the 7-flight fit shown in Fig. 2. 410 The time evolution of BB plume constituents is affected by plume dilution because processes such as gas phase oxidation and the partitioning of POA and SOA between phases are concentration dependent (Hodshire et al, 2019b). In order to provide context for the BBOP data set and to facilitate comparison with other studies we present in Table 4 ambient temperatures, MCE, and peak near-fire values for scattering, mixing ratio of CO and concentration of nr-PM1. A dilution rate for each flight was determined from a fit 415 to the peak mixing ratio of CO on each transect, plotted as a function of photochemical age in Fig. S8. Our measures of dilution are only qualitative as no attempt was made to sample an entire plume cross section.
Within the time taken for photochemical age to change from 0.2 to 1.0, peak plume mixing ratios of CO decreased on average by a factor of 4.3. If measurements had been made at lower altitude, starting closer to the fire, dilution rates would have been much higher (Hodshire et al., 2019b). 420

Flight 730b
The ground track for flight 730b, shown in Fig. 3, consists of two nearly identical sets of transects at 6 downwind distances. Observation times for the second set follow the first set by about one hour. Highest 425 concentrations are on plume-crossing #1; there are short duration spikes with the ratio of NOx to NOy approaching one, and CPC25 concentrations above 10 6 cm -3 . In Fig. 3, plume-crossing #2 appears similar to #1. But on crossing #2 photochemical age has increased, and the CPCs show far fewer particles with diameter below 10 nm. On that basis, it appears that the plume crossings at location #2 are downwind of the main fire region. We cannot, however, dismiss the possibility that the active burning region extends to 430 nominal downwind transects, a consideration pertinent to other flights.
Time series data for nr-PM1, Clˉ, CO, and scattering are displayed in Fig. 4. Peak heights of the conservative tracer CO (and the non-conservative quantities, nr-PM1, Clˉ, and scattering) decrease with distance as the plume becomes wider and more extended in altitude. As downwind distance and 435 photochemical age increase, the traces in Fig. 4b diverge indicating that the ratio of scattering to CO increases. An increase in scattering/CO and MSE with age is a general feature of the wildfires studied.
Transect averaged quantities were used to compare Sets 1 and 2. As examples we present in Fig. 5 photochemical age vs. downwind travel time and the ratio of scattering to nr-PM1 as a function of 440 photochemical age. Differences between Sets 1 and 2 are due to the precision of our measurements and natural time changes in the fire Variations between observations in Set 1 and Set 2 can also be seen in the spatial displacement of plumes between repeated transects (Fig. 3), though allowance should be made for repeated transects not being exactly coincident.

445
In basing our analysis of plume evolution on wide cross-plume averages we are ignoring smaller scale structure. For flight 730b most transects encompass two fire regions separated by 10's of km. One fire region is to the north and one to the south of the dashed line in Fig. 3. A plot of Cl¯ vs. nr-PM1 in Fig.   5 (or a comparison of plots in Fig. 4a) shows that aerosol from the south fire has a Clˉ/nr-PM1 fraction of ~ 2.5%, compared with 0.3% in the north. At the more downwind transects these two plumes overlap giving 450 intermediate ratios. Differences in Clˉ fraction are due to fuel type with the higher Cl¯ fraction characteristic of grass-land fires (Stockwell et al., 2014). Other fire properties vary between north and south. Onasch et al (2018), have shown that the low Clˉ plume has a higher O to C ratio (0.37 vs. 0.31).
The Colockum Tarp fire on flights 730a and 809a and the Mile Marker 28 fire on Flights 725a and 726a also had a bimodal Clˉ to total nr-aerosol ratio. 455

Flight 821b
Concentrations of gasses and aerosols sampled from the Government Flats fire on the 2 nd August 21 flight were a factor of two higher than observed in other wildfires. The ground track, shown in Fig. 7 was a combination of cross-plume traverses and an along-plume segment in which the G-1 travelled against the 460 wind, 45 km toward the fire line.
Time sequences of scattering and nr-PM1 measured on the along-plume flight segment are shown in Fig. 8a. MSE, which is the ratio of these quantities is given in Fig. 8b as a continuous function and averaged over each of 7 transects. The along-plume flight segment starts at 10:04:00 in dilute smoke. 465 Though not readily apparent given the scale of the figure, a plume signature is seen in CO, scattering and nr-PM1 which are 360 ppb, 403 Mm -1 , and 114 μg m -3 , respectively, averaged over transect 7. At 10:06 there is a transition to a higher concentration plume region which encompasses transects 8-13. Between 10:11:50 and 10:11:51, a plume boundary is crossed into much cleaner air. Scattering and nr-PM1 increase between 10:04:00 and 10:11:50, but not monotonically as the path of the G-1 is not consistently oriented 470 along the path of a hypothetical, non-meandering point source plume.
From Fig. 8a, one can see that the ratio of scattering to nr-PM1 increases with downwind distance (decreasing clock time). For the purpose of illustration, Fig. 8b presents 1 Hz values for MSE generated from the smoothed data in Fig 8a (see figure caption). Without smoothing, 1 Hz MSE is too noisy to be 475 useful. In order to bypass such difficulties, we determine transect-averages as in Eqs. 2 and 3. According to the transect-average values in Fig. 8b, MSE increases from 1.9 to 3.6 as plume age changes from 0.12 to 1.0.
The along-plume flight segment of flight 821b provides an illustration of the time dependence of 480 aerosol neutralization, which we describe in terms of a ratio of equivalents: (Clˉ + SO4 2ˉ + NO3ˉ) / NH4 + .
This expression is a simplification as there are other basic anions and acidic cations besides these four measured ions. Also, these anions are not necessarily associated with H + . Fig. 9, shows that near-source aerosol has primary sulfate, nitrate, and chloride that is not matched by the uptake of NH3 resulting in an acidic ion balance over most of the plume. There is a steady trend towards neutralization with downwind 485 distance, with an equivalence ratio of one nearly reached at the furthermost downwind point of the alongplume transect. For almost all of the along-plume segment, the number of equivalents of NH4 + is very nearly the same as the number of equivalents of NO3ˉ, despite the abundance of SO4 2ˉa nd Clˉ, suggesting that much of the sulfate and chloride is in a non-acidic form (Akagi et al., 2012).  Table 5. Changes due to aging are given by (XAged -XFresh)/XFresh, in Table 6.
On the 9 pseudo-Lagrangian flights, aerosol mass normalized by CO, Δnr-PM1/ΔCO, varies between a 22% increase and a 29% decrease. The average change is a 12% decrease with a one σ standard 505 deviation (16%) that encompasses no net aerosol production. Figure 10b shows scattering normalized to CO as a function of photochemical age. According to the summary in Table 6, the aging change in normalized scattering ranges from constant to nearly doubling, with a 9-flight average equal to a 41% increase. As, on average, there is no increase in nr-aerosol mass with age, the normalized scattering increase is not due to more aerosol; rather, it is due to a time evolution in aerosol microphysics. Figure 10c  510 shows the time evolution of scattering per unit of nr-aerosol mass. MSE increases with photochemical age for all 9 flights; the average increase is 56% with a standard deviation of 20% and range 33 to 97%. Figure 11c indicates that on 7 of 9 flights, absorption per unit mass of aerosol is either independent of-or decreases slightly with age. The increase observed for flight 809a appears to be within the scatter of data points. A nearly constant absorption is consistent with results presented by Sedlacek et al (2018a) and much like the observations of Forrister et al (2015) which extend to longer time periods. Changes in SSA with atmospheric processing (Fig. 11d) are therefore controlled by the increase in scattering as the plume ages (Fig. 10c). The lowest SSAs observed were 0.8 to 0.85 in fresh smoke. As the plume ages for 2 -3 hours, SSAs are 0.9 to 0.98. 520

Aerosol composition
Age related changes in the chemical composition of BB aerosol are presented here in terms of the fractional contribution of black carbon, organic compounds and inorganic ions (ΔNH4 + , ΔCl -, ΔNO3 -. and

rBC and inorganics
Aerosol composition measurements for fresh and aged BB aerosol are summarized in Fig. 12. In every transect, more than 90% of aerosol mass is OA. Flight-averaged rBC is between 0.5 and 2.5% of nr-PM1. On average, the mass fraction of rBC in aged smoke is 11% higher than in fresh emissions (Fig. 12). 535 This increase is consistent with a constant amount of rBC as the concentration of nr-PM1 decreases by very nearly the same amount (12%, Table 6) relative to CO. Aging yields a change in ΔrBC/ΔCO of 0.5%.
There is a one-σ standard deviation of 21%, which is due to uncertainty of our methodology (i.e. fire inhomogeneities, uncertain backgrounds) and measurement uncertainties in CO and rBC. Inorganic species constitute less than 10% of nr-PM1. Differences between fires are comparable to the differences between 540 fresh and aged emissions. In the 730b case study Clˉ differed by about an order of magnitude according to fuel type. It is likely that primary Clˉ is in the form of non-volatile KCl and the decrease in Clˉ reflects the formation and volatility of HCl (Akagi et al., 2012). On most flights SO4 2ˉ decreases with age but the correlations are poor. An increase is expected from oxidation of SO2 (Yokelson et al., 2009;Akagi et al, 2012). SO2 emission factors measured on flights 730a and 730b are given by Liu et al (2016). NO3¯ and 545 NH4 + increase with age (Fig. S9). Increases in NO3ˉ would occur from the formation and partitioning of HNO3 to the aerosol phase. Fig. 12 shows that the 9-flight average acidity follows the aging pattern seen on flight 821b (Fig.   9b). Fresh emissions are acidic with NO3ˉ equivalent to NH4 + , despite the presence of SO4 2ˉ and Clˉ. The 550 same caveats apply regarding a description of acidity in terms of the 4 ions readily quantified by the AMS.

Organic aerosol
The speciation of organic compounds is expected to vary with age as primary compounds are lost by evaporation or transformed by aerosol phase chemistry. The later pathway includes the transformations 555 that convert primary organics to TBs (Adachi et al., 2019). As a plumes ages, gas phase oxidation of VOCs creates less-volatile products which partition to the aerosol phase as SOA. Though organic mass is on average constant in BBOP, the constituent species become more oxidized and viscous with age. Transectaverage H to C and O to C ratios (H:C, and O:C, respectively) observed in 9 wildfire flights are depicted by a Van Krevelen plot in Fig. 13. Isolines of OSc, the average carbon oxidation state, defined to good 560 approximation as OSc = 2 O:C -H:C (Kroll et al., 2011), indicate, in each flight, an increase in carbon oxidation state with photochemical age. For the 9 BBOP flights, slopes in the Van Krevelen plots vary from near zero to -1 (average value = -0.4), implying flight to flight variability in the predominant oxidation mechanism and products. For example, aerosol would age with a slope of zero if oxidation proceeded by adding OH or OOH moieties; a slope of -1/2 by formation of a carboxylic acid with 565 fragmentation and a slope of -1 by the addition of a carboxylic acid without fragmentation. Adachi et al (2019) used scanning transmission X-ray spectroscopy and electron energy loss spectrometry analysis to show that the formation of TBs is accompanied by an increase in carboxylic acids.
A comparison with other BB aerosol elemental ratios (Ng et al., 2016), shows that BBOP elemental 570 ratios have values characteristic of semi-volatile OOA species and that the 2 to 3.5 hours of aging observed on BBOP flights leads to composition changes in the direction of lower volatility species. A similar conclusion that BBOP aerosol is in the semi-volatile category, progressing to lower volatility with age is supported by values of f43 and f44 shown via a "triangle plot" in Fig. 14. The age dependence of f43, f44, f60, H:C, O:C, and OSc is given explicitly in Fig. S10. Quantities directly related to aerosol oxidation state, 575 f44, O:C and OSc monotonically change with age for all flights, albeit with some scatter and flight-to-flight differences. f60 which is a surrogate for primary levoglucosan, and related compounds (Cubison et al., 2011) decrease with age for all flights (Fig. S10c).
Measurements from two regional scale BBOP flights, 806a and 816a, (not part of the current study) 580 have been compared with observations at MBO. According to back-trajectories, MBO sampled in the same air masses as the G-1 after 6-12 hours additional processing (Collier et al., 2016;Zhou et al., 2017;Zhang et al., 2018). A PMF analysis of the MBO observations (Zhang et al., 2017)  Section 5 Increase in MSE with age

Observations 595
Transect-average measurements in Fig. 10c and summarized in Table 6, indicate an average increase in mass scattering efficiency (MSE) of 56% in aged as compared with fresh aerosol. For each transect a surrogate for particle size is determined from FIMS measurements as follows: Size spectra are first averaged over a transect. The peak in the transect-averaged size spectrum is determined from a local quadratic fit that uses the 3 to 5 size bins nearest the largest value of dN/dLogDp. A restriction to Dp > 100 600 nm yields a peak for dN/dLogDp within the accumulation mode. We will refer to this diameter as DGEO, in analogy with the geometric mean of a log normal distribution. Fig. 15 shows that DGEO is an increasing function of photochemical age in accord with multiple field observations of BB aerosol (e.g. Akagi et al., 2012;Eck et al., 2013;Carrico et al., 2016). MSE is 605 observed to increase with DGEO (Fig. 16). This trend is expected for size distributions dominated by  (2016), the ensemble of flights shows little 610 relation between MSE and DGEO. The low correlation between MSE and DGEO for the BBOP 9-flight ensemble appears to be caused by real fire-to-fire differences; said differences might be diminished by additional atmospheric processing in the more aged smoke seen at MBO by Liang et al (2016).
In the size range of our samples, backscatter ratios and scattering Angstrom exponents decrease 615 with particle size (Selimovic et al., 2019). An anti-correlation between these intensive parameters and MSE is expected and is observed (Figs. 17-18), both for individual flights and for the ensemble of 9 flights that cover a range of MSE from 1.5 to 6.
An increase in MSE could be caused by an increase in the real part of the refractive index, mR. In 620 biomass burn aerosol generated in the laboratory or sampled in the ambient atmosphere, mR is observed to be ~ 1.50 to 1.60 (Levin et al., 2010;Adler et al., 2011) with values above 1.60 occurring in smoke that has a higher fraction of BC than seen in BBOP. Based on literature values and Sedlacek et al (2018a), we estimate that wildfire generated aerosol in BBOP has a refractive index, m = 1.53 -0.02i. A change in mR from 1.53 to 1.60 during plume aging would result in a 17 to 24% increase in MSE according to Mie calculations of aerosol with a log normal distribution in which DGEO = 200 nm and σG (geometric mean standard deviation) between 1.4 and 1.8. However, there are reports that mR is insensitive to aging (Levin et al., 2010) or decreases in aged aerosol (Adler et al., 2011).

Coagulation Calculations 630
An increase in MSE with age is more likely to be caused by a rearrangement of particle mass, favoring large diameter efficient scatters at the expense of small inefficient scatters. An example calculation shows that coagulation can lead to a substantial (factor of 2) increase in MSE.
Effects of coagulation upon an initial aerosol size distribution (Sakamoto et al., 2016) were 635 determined using a Brownian coagulation kernel and algorithms in Jacobson et al (1994). Calculations were initialized with the near-fire FIMS particle size distribution measured on transect 13 of the 821b flight and integrated for 6600 sec, corresponding to the estimated atmospheric transport time on the along-plume flight segment, between the fire and the furthest downwind sample on transect 7 (see Fig. 7). Particles with diameters above 260 nm are qualitatively accounted for by fitting the FIMS measurements to a double log 640 normal that extends to 1000 nm. Based on CO measurements, the simulated air mass was diluted by a factor of 4 between transects 13 and 7. Particle concentrations in background air are negligible compared with the plume and are ignored. The dilution rate is an average value, so it is not expected that calculated concentration will follow the increases and decreases observed as the G-1 encountered more and less concentrated areas within the plume. Errors in calculated aerosol concentration effect rates of coagulation 645 but do not directly affect the aerosol size distribution shape or MSE which are intensive quantities.
Transect-averaged aerosol size distributions, extrapolated to 1000 nm, for the along-plume segment of flight 821a are given in Fig. 19 along with the corresponding size distributions determined from coagulation calculations. The near-source size distributions have peaks in the Aitken and accumulation 650 mode size range. The FIMS and calculated size distributions show a growth in the number of accumulation mode particles relative to the Aitken mode as well as a shift in the accumulation mode to larger particle diameters as time increases. Coagulation calculations produce changes comparable to the FIMS measurements between transects 13 and 10. Further downwind, the transfer of material from Aitken to accumulation mode size ranges is more pronounced in the FIMS observations than in the coagulation 655 calculations. MSEs calculated from the observed and calculated aerosol size distributions are compared in Correlations found in BBOP flights between i) MSE and ii) DGEO, aerosol Angstrom exponent, and aerosol backscatter ratio are evidence that increases in MSE with time are due to the growth of aerosol particles, such that they become more efficient scatters. Similar correlations have been found in other 665 biomass burn studies (Levin et al., 2010;Liang et al., 2016;Selimovic et al., 2019). Coagulation was shown to be one mechanism for redistributing mass between different size particles that can lead to a doubling of MSE. Because the BBOP data set is restricted to aerosol size measurements below 210 or 260 nm, it is difficult to quantify whether coagulation alone can reproduce the observed MSE age dependence.

670
Coagulation is certainly not the only aerosol microphysical process occurring as evidenced by aerosol chemical composition changes with age (Fig. S10). Chemical changes in the organic component of BB aerosol have been interpreted in terms of mass transfer between particulate and gas phases. In this scenario, high volatility POA evaporates as the plume dilutes. Gas phase reactions of primary flame emissions or the evaporated POA, creates oxygenated lower volatility VOCs that subsequently partition as 675 SOA to the aerosol phase (Grieshop et al., 2009;Hennigan et al., 2012;Jolleys et al., 2012;May et al., 2013;Morgan et al., 2019). Transfer of material between the gas and (bulk) aerosol phase and amongst particles in the aerosol phase is driven by the thermodynamic constraint of equalizing chemical potentials.
Though thermodynamics defines quasi-stationary states in an evolving plume, actual distributions are commonly dictated by the kinetics of mass transfer within particles, between phases, and between 680 individual particles (Marcolli et al., 2004;Zhang et al., 2012) Time-dependent changes in the chemical composition of wildfire plumes observed in BBOP are consistent with a decrease in POA compensated by an increase in SOA, such that the total concentration of organic aerosol, corrected for dilution, is approximately constant. An increase of SOA is inferred by the 685 systematic increase in f44, O to C ratio and carbon oxidation state with time (Figs, S10b, e, and f). A decrease in f60 (Fig. S10c) has been linked to the evaporation of primary levoglucosan and related compounds.
Changes in chemical markers used to characterize the carbon oxidation state are several-fold 690 smaller than observed in laboratory burns and in most field samples. Thus, the BBOP data set occupies a small fraction of composition space depicted in "triangle" diagrams of f44 vs f43 (Ng et al., 2011) and a similarly small fraction of a Van Krevelen diagram (Heald et al., 2010). Arrows that one could construct between our fresh and aged samples point towards the more aged compositions that others have observed (se Figs. 1 and 4, Ng et al., 2011). Most of the plume samples described in this study have been exposed to 695 atmospheric processing for two hours or less; the oldest sample are exposed for less than 3.5 hours. In contrast the data sets used in Heald et al. (2010) and Ng et al. (2011) contain both fresh and more processed BB smoke with atmospheric residence times (or equivalent OH exposure) of days.
Given that there are aged-related changes in organic composition (though relatively small) 700 consistent with loss of POA and gain of SOA, is it possible that a transfer of mass between the gas and particulate phase and amongst particles can yield a size distribution in which MSE increases by as much as a factor of two? It is likely that during mass exchange between the gas and aerosol phase, equilibrium amongst particles of different sizes is not maintained since the air mass is evolving rapidly (i.e. <1 hour) and mass transfer between particles can have substantially longer time constants (Marcolli et al., 2004). 705 The kinetics of diffusion in the continuum regime tends to favor evaporation of small particles since in that regime, dDp/dt ~ Dp -1 . A preferential evaporation of small particles occurs also for the transition regime size particles that produce most of the scattering in wildfire plumes. As total mass is nearly invariant, the concurrent process of condensation of SOA cannot have the same kinetics as evaporation of POA, else particles of a given size will have no net size change. If SOA was to be distributed amongst particles so as 710 to equalize mole fractions (i.e. follow a volume growth law), then growth of large particles would be favored over smaller ones and MSE would increase. Detailed calculations invoking a wider choice of evaporation and growth options as well as possible kinetic limitations caused by within-aerosol diffusion are required to quantify effects of mass transfer on MSE.

Conclusions
As part of the BBOP campaign in the Pacific NW U,S., nine wildfire plumes were sampled in a pseudo-Lagrangian mode to determine the time evolution of pollutants between the fire and after up-to 3.5 hours of daytime atmospheric processing. Atmospheric processing was quantified by a photochemical age defined as -Log10(NOx/NOy). Typical ages were between 0.2 and 1.0 and maximum downwind times 720 close to 2 hours, though on occasion as long as 3.5 hours. Plume concentrations were corrected for dilution using CO as a conservative tracer. Background subtraction depended on observations in near-by clean air.
On average, normalized aerosol mass concentrations were constant over several hours of atmospheric aging. Mass scattering efficiency increased with age by an average of 56%; the range amongst 725 flights was an increase of 33% to 97%. Except for two flights, mass absorption coefficients (MACs) are in the high teens and nearly independent of age. If absorption is due to coated BC, then coatings are formed early in the BB plume and of a thickness such that absorption is insensitive to further coating (Bond et al, 2006;Forrister et al., 2015). Scattering, normalized for dilution, increases with age causing SSA to likewise increase. In fresh smoke albedos were 0.85 -0.90. These SSAs increased to ~ 0.95 in aged smoke. 730 If we suppose our plumes to be over a portion of the globe with average albedo (that average including oceans, etc.), then the wildfire plumes when first observed would have a near-zero direct radiative effect on the Earth's radiative balance and a cooling effect after 2 -3 hours (Selimovic et al, 2019;Eck et al 2013). For typical BB particles, scattering and MSE are increasing functions of particle size. In order to 735 establish that aerosol particles grow with age we relied on i) a decrease in aerosol backscatter ratio and Angstrom exponent with age (as expected from Mie theory) and ii) an increase with age of DGEO, the diameter, within the accumulation mode, at which point dN/dLogDp has a maximum value. DGEO is best thought of as a surrogate of particle size as it is obtained from measurements with Dp < 210 or 260 nm, short of covering the full range of optically active particles. Almost all values for the surrogate measures of 740 Dp increase with age. Fire to fire variability of these metrics can be larger than the difference between fresh and aged emissions, which cautions against comparing fresh emissions from one fire with aged emissions from another.
A calculation from a high concentration along-plume flight segment indicates that coagulation can 745 cause a factor of two increase in MSE by the transfer of mass from the Aitken mode to the larger accumulation mode. Further calculations constrained by aerosol size distributions that fully cover the size range that contributes most to scattering, are required to determine the importance of coagulation and to identify other mechanisms that cause MSE to increase with age.

750
In newly emitted plumes NH4 + and NO3ˉ tended to have an equivalence ratio near unity, despite the presence of Clˉ and SO4 2ˉ. That HNO3 and HCl were not volatilized argues against the initial SO4 2ˉ and Clˉ being cations of strong acids. Primary Clˉ varied between 0.2 and 2.5%, with the higher value associated with grass fuels. Aged plumes had, on average, a neutral equivalence ratio (anions / NH4 + near unity). The ratio, rBC/nr-PM1 varied amongst flights, with a range of 0.5% to 2.5%. In all cases organic 755 aerosol constituted > 90% of nr-PM1.
Organic composition changed with age as POA evaporated and SOA condensed. Loss of POA can be inferred from the age dependence of f60, a surrogate for primary emissions of levoglucosan and related compounds. Condensation of SOA is seen from the increase in f44 and O to C ratio in aged samples. 760 Changes with age of these mass spectra fragments, though robust, are smaller than that observed in most laboratory and field studies of aerosol aging. This is expected as our aged samples were typically only2 to 3 hours downwind of their wildfire source. Data Availability. Data from the BBOP field campaign have been archived at https://www.arm.gov  Data from these probes were not used because of severe coincidence problems. § § Upper size limit during plume sampling = 210 nm for flights 809a and 822a; 260 nm for flights 25a, 726a, 730a, 730b, 814a, and 822b.
D Instrument sampled from dilution line.   a Photochemical age determined from downwind distance and 7-flight fitting function shown in Fig. 2.
b ΔCO2 > 2 ppm, (±1 σ for variation between transects)  with data, aging causes the difference between the SSA of fresh smoke and unity to decrease by more than a factor of 2.   Along-plume transects, 7-13, labelled in red. Transect 11 had limited NOx data and is not included in graphs in which photochemical age is the independent variable. Latitude and Longitude not to scale.