The Importance of Size Ranges in Aerosol Instrument Intercomparisons:

Aerosol intercomparisons are inherently complex, as they convolve instrument-dependent 12 detection efficiencies vs. size (which often change with pressure, temperature, or humidity) and variations 13 on the sampled aerosol population, in addition to differences in chemical detection principles (e.g., 14 including inorganic-only nitrate vs. inorganic plus organic nitrate for two instruments). The NASA 15 Atmospheric Tomography Mission (ATom) spanned four separate aircraft deployments, which sampled 16 the remote marine troposphere from 86°S to 82°N over different seasons with a wide range of aerosol 17 concentrations and compositions. Aerosols were quantified with a set of carefully characterized and 18 calibrated instruments, some based on particle sizing and some on composition measurements. This study 19 aims to provide a critical evaluation of the size-related factors impacting aerosol intercomparisons, and 20 of aerosol quantification during ATom, with a focus on the Aerosol Mass Spectrometer (AMS). The 21 volume determined from physical sizing instruments is compared in detail with that derived from the 22 chemical measurements of the AMS and the Single Particle Soot Photometer (SP2). Special attention was 23 paid to characterize the upper end of the AMS size-dependent transmission with in-field calibrations, 24 which we show to be critical for accurate comparisons across instruments with inevitably different size 25 cuts. Observed differences between campaigns emphasize the importance of characterizing AMS 26 transmission for each instrument and field study for meaningful interpretation of instrument comparisons. 27 Good agreement was found between the composition-based volume (including AMS-quantified sea salt) 28 and that derived from the size spectrometers. The very clean conditions during most of ATom resulted in 29 substantial statistical noise (i.e., precision error), which we show to be substantially reduced by averaging 30 at several-minute time intervals. The AMS captured, on average, 95 ± 15% of the standard PM1 volume. 31 These results support the absence of significant unknown biases and the appropriateness of the accuracy 32 estimates for AMS total mass/volume for the mostly aged air masses encountered in ATom. The particle 33 size ranges that contribute chemical composition information to the AMS and complementary 34 composition instruments are investigated, to inform their use in future studies. 35 https://doi.org/10.5194/amt-2020-224 Preprint. Discussion started: 15 June 2020 c © Author(s) 2020. CC BY 4.0 License.

concentrations and compositions. Aerosols were quantified with a set of carefully characterized and 18 calibrated instruments, some based on particle sizing and some on composition measurements. This study 19 aims to provide a critical evaluation of the size-related factors impacting aerosol intercomparisons, and 20 of aerosol quantification during ATom, with a focus on the Aerosol Mass Spectrometer (AMS). The 21 volume determined from physical sizing instruments is compared in detail with that derived from the 22 chemical measurements of the AMS and the Single Particle Soot Photometer (SP2). Special attention was 23 paid to characterize the upper end of the AMS size-dependent transmission with in-field calibrations, 24 which we show to be critical for accurate comparisons across instruments with inevitably different size 25 cuts. Observed differences between campaigns emphasize the importance of characterizing AMS 26 transmission for each instrument and field study for meaningful interpretation of instrument comparisons. 27 Good agreement was found between the composition-based volume (including AMS-quantified sea salt) 28 and that derived from the size spectrometers. The very clean conditions during most of ATom resulted in 29 substantial statistical noise (i.e., precision error), which we show to be substantially reduced by averaging 30 at several-minute time intervals. The AMS captured, on average, 95 ± 15% of the standard PM1 volume. 31 These results support the absence of significant unknown biases and the appropriateness of the accuracy 32 estimates for AMS total mass/volume for the mostly aged air masses encountered in ATom. The particle 33 size ranges that contribute chemical composition information to the AMS and complementary 34 composition instruments are investigated, to inform their use in future studies.

36
Aerosols are ubiquitous in the atmosphere and have a lifetime of about a week, and thus can travel 37 long distances (Tsigaridis et al., 2014), and have important effects on climate forcing, through both direct 38 (Pilinis et al., 1995;Haywood and Boucher, 2000) and indirect effects (Lohmann and Feichter, 2005;39 IPCC, 2013). Remote regions account for much of the Earth's surface and are infrequently sampled, and 40 thus have especially uncertain aerosol distributions and radiative impacts (IPCC, 2013;Hodzic et al., 41 2020). The NASA Atmospheric Tomography Mission (ATom) sampled the remote marine troposphere 42 from 86°S to 82°N over four different seasons with a comprehensive suite of high-quality and carefully 43 calibrated and operated physical and chemical aerosol instruments. It provides a unique dataset to improve 44 our understanding of the remote atmospheric aerosols and thus refine global model predictions. A 45 prerequisite for that purpose is to evaluate the accuracy and consistency of the ATom aerosol instruments. 46 The ATom physical sizing instruments have been recently described and evaluated in Williamson 47 et al. (2018), Kupc et al. (2018), and Brock et al. (2019), while the Particle Analysis by Laser Mass 48 Spectrometer (PALMS) chemical instrument during ATom has been described in Froyd et al. (2019). In 49 this paper, we focus on the Aerodyne Aerosol Mass Spectrometer (AMS). AMS (Canagaratna et al., 2007) 50 and Aerosol Chemical Speciation Monitor (ACSM, smaller, lower cost, and simpler to operate versions) 51 (Ng et al., 2011), have been deployed extensively worldwide for ground aerosol monitoring (Jimenez et 52 al., 2009;Crenn et al., 2015;Hu et al., 2015;Kiendler-Scharr et al., 2016;Zhang et al., 2018;ACTRiS, 53 2019). AMS has been deployed in most advanced atmospheric chemistry aircraft experiments worldwide 54 (Dunlea et al., 2009;Middlebrook et al., 2012;Barth et al., 2015;Schroder et al., 2018;Garofalo et al., AMS for remote aerosols, and (3) the size ranges contributing chemical composition information to 64 different instruments for ATom, and their variation with altitude. Volume comparisons probe the ability 65 of the AMS to quantify total mass and predict aerosol density based on fractional composition accurately, 66 and hence is the most germane comparison for total quantification. We examine in detail the accurate 67 quantification and application of the AMS transmission efficiency (EL) to the particle volume 68 intercomparisons in this study. This study also serves as the basis for a future study on individual chemical 69 species intercomparisons. repeatedly ascended and descended between ~0.18 and ~13 km altitudes at regular intervals, typically 78 every hour (with a single vertical profile lasting ~25 min), leading to executing ~140 vertical profiles of 79 the troposphere per deployment. The unique spatio-temporal coverage and high-quality measurements of 80 this campaign ensure that its data will be used very widely, such as to evaluate and constrain global 81 modeling. Therefore it is of high interest to document the consistency of the multiple aerosol 82 measurements. This analysis is also useful to re-evaluate the quantification uncertainties of the AMS for 83 a wide range of particle concentrations and composition (e.g., Fig. S1 in the supplementary info, SI). Due 84 to the similarities in the geographic coverage of ATom studies, we focus on the intercomparisons for the 85 first two ATom campaigns in the following analysis. 86 2.2 Definitions of particle diameters 87 Conversions between different particle diameter definitions are required for meaningful 88 instrument comparisons. For example, particle size spectrometers report estimated geometric diameter (dp), which is derived from multiple condensation particle counters using an inversion method, or from 90 light scattering signals by using an assumed constant refractive index for aerosols. AMS transmission 91 operates in vacuum aerodynamic diameter (dva) since its aerodynamic lens and supersonic expansion 92 operate in the free molecular regime (DeCarlo et al., 2004). Impactors (Marple et al., 1991(Marple et al., , 2014 and 93 cyclones (typically sourced from URG Corp., Chapel Hill, NC, USA) are often installed upstream of 94 aerosol instruments to preselect desired aerosol ranges for ground or aircraft measurements. The cutoff 95 sizes of both devices follow the transition-regime aerodynamic diameter (dta; as the size range of interest 96 to this study is in the transition regime, requiring a "slip correction"). A detailed discussion of particle 97 diameters definitions can be found in DeCarlo et al. (2004). dva is related to the volume-equivalent 98 diameter (dve, the diameter that would result if the particle was melted to form a sphere of the same density 99 as the particle and without any internal voids) as: is the particle density, 0 is the standard density (1 g cm -3 ), and is the vacuum (i.e., free-101 molecular regime) dynamic shape factor (=1 for spheres and >1 for non-spherical particles). Since the 102 aerosols sampled during ATom were remote and aged, we assume ~1 and ~. The transition-103 regime aerodynamic diameter can be calculated as: where is the transition-regime dynamic shape factor, and is the Cunningham slip correction factor. 105 In this study, is assumed as 1 and is calculated based on air pressure. Although a given particle

111
The highly customized University of Colorado (CU) high-resolution time-of-flight aerosol mass 112 spectrometer (HR-ToF-AMS, hereafter referred to as AMS; Aerodyne Research Inc., Billerica, MA) 113 (DeCarlo et al., 2006) measured non-refractory, bulk submicron particles composition at 1 Hz resolution. 114 The AMS uses an aerodynamic lens to sample particles into a high vacuum, where they impact and 115 vaporize on a hot porous tungsten vaporizer (600 °C). The evaporated constituents undergo electron 116 ionization (EI), with the resulting ions being detected by a mass spectrometer (Jayne et al., 2000;Jimenez 117 et al., 2003;Drewnick et al., 2005;DeCarlo et al., 2006;Canagaratna et al., 2007). The mass concentration 118 of a species, s, within a multi-component aerosol particle can be calculated from the measured ion signal 119 with the following equation (Alfarra et al., 2004;Canagaratna et al., 2007;Jimenez et al., 2016): where Cs is the mass concentration of species s, MWNO3 is the molecular weight of nitrate, CEs is the 121 collection efficiency of species s, RIEs is the relative ionization efficiency of species s (to nitrate), IENO3 122 is the ionization efficiency of nitrate, Q is the volume flow rate into the AMS, NA is Avogadro's number, 123 Is,i is the ion signal from ion i produced from species s, and the 10 12 factor accounts for unit conversions.

124
CE is typically defined as the efficiency with which particles entering the AMS inlet are detected. 125 It has been formally defined as a product of aerodynamic lens transmission efficiency for spherical 126 particles (EL), transmission efficiency correction for non-spherical particles (Es) due to additional particle 127 beam broadening, and detection efficiency at the vaporizer (Eb), which can be reduced due to particle 128 bounce. It is thus expressed as 129 = × × (4) (Huffman et al., 2005;Canagaratna et al., 2007;Middlebrook et al., 2012). Previous studies have shown quantified as it is experimentally challenging to do so. Eb depends on particle viscosity and thus phase 135 (Matthew et al., 2008;Middlebrook et al., 2012;Pajunoja et al., 2016). With the "standard vaporizer" 136 used in this study (Hu et al., 2020), ambient aerosols in continental regions typically have Eb~0.5, but a 137 range between 0.5 to 1 can be observed (Middlebrook et al., 2012;Hu et al., 2017Hu et al., , 2020. Eb increases 138 for certain compositions that lead to less viscous particles, such as high ammonium nitrate mass fraction 139 or high acidity conditions, which can be estimated with a parameterization based on aerosol composition 140 (Middlebrook et al., 2012;Hu et al., 2017Hu et al., , 2020Nault et al., 2018). Such parametrizations assume 141 internally mixed aerosols, which is typically the case for submicron ambient aerosol away from sources 142 due to condensation and coagulation (Petters et al., 2006;Wang et al., 2010;Mei et al., 2013). CE is 143 estimated to contribute substantially to the overall uncertainty of AMS concentration measurements 144 (Bahreini et al., 2009). 145 The main submicron inorganic ambient aerosol species are ammonium (NH4), sulfate (SO4), 146 nitrate (pNO3), and chloride (Chl), and in marine areas, sea salt. The charges are omitted for the AMS-147 measured nominally inorganic species, as the AMS may also detect some SO4 or NO3 signals from 148 organosulfates or organonitrates (Farmer et al., 2010). To avoid the confusion between the NO3 radical 149 and particle NO3, pNO3 is used to denote total particle NO3 explicitly (Nault et al., 2018). RIEs for the 150 inorganic species can be calibrated regularly (including in the field). However, similar explicit 151 calibrations cannot be readily performed for the thousands of individual organic aerosol (OA) molecules 152 in ambient particles. Thus, laboratory-based calibrations with a limited set of OA species have been used 153 to estimate RIEOA (Slowik et al., 2004;Dzepina et al., 2007;Jimenez et al., 2016;Robinson et al., 2017;154 Xu et al., 2018), and this approach has been verified using laboratory and field intercomparisons with 155 other instruments (Takegawa et al., 2005;Dzepina et al., 2007;DeCarlo et al., 2008;Bahreini et al., 2009;156 Dunlea et al., 2009;Timonen et al., 2010;Docherty et al., 2011;Middlebrook et al., 2012;Crenn et al., for RIEOA of chemically-reduced species such as hydrocarbons, with some values around 1.4 and others 163 higher (Slowik et al., 2004;Dzepina et al., 2007;Docherty et al., 2011;Jimenez et al., 2016;Reyes-164 Villegas et al., 2018;Xu et al., 2018). However, such species were insignificant during ATom. For more 165 oxidized species, relevant to most biomass burning OA and secondary organic aerosol (SOA), average 166 laboratory RIEOA overlaps within uncertainties of 1.4 (Jimenez et al., 2016;Xu et al., 2018). Reviews on 167 this topic (Jimenez et al., 2016;Murphy, 2016aMurphy, , 2016b

173
The aircraft operation of the CU AMS has been discussed previously (DeCarlo et al., 2006(DeCarlo et al., , 2008(DeCarlo et al., , 174 2010Dunlea et al., 2009;Cubison et al., 2011;Kimmel et al., 2011;Schroder et al., 2018 (Vay et al., 2003). Aerosols were introduced at a constant standard flow rate of 9 sL min -184 1 (up to ~9 km, 15 L min -1 above that; "s" refers to standard conditions, and no "s" indicates a volumetric 185 flow at in-situ T and P), with 1 L min -1 being continuously subsampled into a pressure controlled inlet 186 (PCI) operated at 250 mbar (187 Torr) (Bahreini et al., 2008). A fraction of that flow, 94 scm 3 min -1 , was 187 then sampled into the high vacuum region of the mass spectrometer through an aerodynamic focusing 188 lens operated at 2.00 mbar (1.50 Torr). Due to the much lower ambient air pressure at high altitudes, the  (Schroder et al., 2018), also provided multiple AMS transmission measurements throughout the 213 campaign, by a direct comparison of the single-particle AMS counts with a Condensation Particle Counter 214 (CPC) (Nault et al., 2018). Besides this single-size (at the edge of the Another concern for airborne sampling with an AMS is the misalignment of the aerodynamic lens 229 due to mechanical stress during flight. Such a misalignment will not necessarily be caught by the 230 previously described calibrations, since they do not probe the full surface of the vaporizer, and since lens 231 focusing can have some size-dependence. Hence for ATom 2-4, a particle beam width probe (Huffman 232 et al., 2005) was flown and profiles of both the air and particle signal were taken at most airports during 233 the mission, as shown in Fig. S9, directly confirming the lack of change in lens alignment. 234 During ATom, the AMS was operated in the fast mass spectrum mode (Kimmel et al., 2011), 235 allowing for high-time-resolution measurements at 1 Hz. For every minute, AMS started with fast mass 236 spectrum mode with the particle beam blocked (instrumental background measurement; 6 s) and then 237 with the beam open (background plus ambient air and particles; 46 s) and ended with efficient particle min product, the raw mass spectra were averaged prior to data reduction and analysis, which reduces 246 nonlinear spectral fitting noise for the least-squares error minimization method. This is observed because 247 a fit to the 1 min average spectrum has less fitting noise than the average of the fits to the 1 s spectra. In 248 the following analysis, the 1 min data product is used due to their improved signal-to-noise ratio (SNR). 249 Since the aerosol loadings were typically low and changed slowly in the global remote regions, longer  ratios found for concentrations close to the DL). The "improved-ambient" method was used for OA 273 elemental analysis (Canagaratna et al., 2015;Hu et al., 2018). The combined density of SO4, NH4, and 274 pNO3 is assumed as 1.75 g cm -3 , an approximation from ammonium sulfate, ammonium bisulfate, and 275 ammonium nitrate (Sloane et al., 1991;Stein et al., 1994;Salcedo et al., 2006). The non-refractory 276 chloride density is assumed as 1.52 g cm -3 based on ammonium chloride (Salcedo et al., 2006). The 277 frequency distributions of and are summarized in Fig. S12. The mass-weighted average is 278 1.60 ± 0.14 g cm -3 and 1.70 ± 0.10 g cm -3 , and (averaged from above the concentrations above OA 279 DL) is 1.51 ± 0.19 g cm -3 and 1.59 ± 0.24 g cm -3 for ATom-1 and ATom-2, respectively. Negative AMS

285
The following instruments all sampled through the LARGE inlet, except Soluble Acidic Gases 286 and Aerosol (SAGA). The transmission efficiency for this inlet has been characterized as a function of 287 particle size by flying the NASA DC-8 in a previous campaign (McNaughton et al., 2007), demonstrating 288 a unity efficiency up to supermicron size ranges and reaching 50% at dta,air of ~5 µm at the surface and 289 3.2 µm at 12 km. Hereafter, we refer to the 50% transmission diameter as d50.

290
Particle size spectrometers: Dry particle size distributions for dp from 2.7 nm to 4.8 µm were 291 reported at 1 Hz using three optical particle spectrometers, including a Nucleation-Mode Aerosol Size merge the three non-thermally denuded size distributions into one. Hereafter, we refer to the non-300 thermally denuded integrated volume (2.7 nm-4.8 µm) as the physical sizing-based volume (Vphys). AMP 301 performed well during ATom. Most relevant to the AMS size range, the UHSAS reported volume was 302 estimated to have an asymmetric uncertainty of +12.4%/-27.5% due to the differences in refractive index 303 (n) between ambient particles and assumed ammonium sulfate particles (n = 1.527, which is similar to 304 the refractive index found for aged ambient OA (Aldhaif et al., 2018)). This uncertainty range is estimated 305 to be between 1σ and 2σ depending on the conditions. Here we assume that it represents 1.5σ when using 306 it for uncertainty analyses. The ATom SP2 detection system was operated as in Schwarz et al. (2010a) with a size range for rBC 311 mass of dve ~90-550 nm (Schwarz et al., 2010b). This size range typically contains ~90% of the total rBC 312 mass in the ambient accumulation mode (Schwarz et al., 2008;Shiraiwa et al., 2008).

PALMS:
The Particle Analysis by Laser Mass Spectrometry (PALMS) is a single-particle laser-314 ablation/ionization mass spectrometer instrument that measures size-resolved (dp ~ 0.1-5 µm) particle 315 chemical composition with fast response (Thomson et al., 2000;Murphy et al., 2006). Particle mass 316 concentrations can be derived as a function of size when mapping the PALMS chemical composition to 317 the size distributions reported from the UHSAS and LAS, which is referred to as the PALMS-AMP 318 products (Froyd et al., 2019). In this study, we focus on the different particle size ranges observed by 319 PALMS and AMS, to illustrate the strengths and applications of the two aerosol composition instruments 320 onboard the DC-8. 321 PALMS is the most complex of the chemical composition instruments used in ATom. It is a single-322 particle based instrument with both a very steep detection efficiency vs. particle size in the smaller particle 323 range and the ability to measure much larger particles than the AMS. While the total reported mass (with 324 some density uncertainty) of the PALMS-AMP products will always match the physical volume 325 measurement over the range that PALMS reports (100-5000 nm dp), the uneven sampling data coverage of particles across each size bin, as well as the broadness of the bins chosen for PALMS-AMP analysis, 327 can lead to a chemical bias if composition gradients exist within a bin (Fig. S13). Therefore, care must be 328 taken to balance statistical representativeness against the need for unvarying particle composition across 329 the size range over which those statistics are obtained (Froyd et al., 2019). 330 For intercomparisons we characterize the specific size range over which the PALMS can obtain 331 sufficient chemical information over a given time period under the ATom conditions, which is mainly 332 limited by particle statistics. If zero or a very low number of particles is sampled for a given AMP size 333 bin and time period, there is no real information being captured for characterizing the composition of the 334 particles in that bin. That is true even if the AMP volume in that bin is assigned a composition by 335 extrapolating the composition of larger or smaller particles. Therefore, we derived the PALMS detected 336 particle numbers based on the raw AMP size resolution (20 bins/decade, 34 bins in total above 100 nm dp 337 for the size range that PALMS-AMP reports) to avoid the assumption of homogeneous chemical 338 composition within four broader bins in Froyd et al. (2019). This provides an alternative illustration of 339 PALMS size coverage and introduces a method that is applicable to other single-particle mass 340 spectrometers or other particle-counting based chemical instruments. A sensitivity test was carried out at , we assume that if PALMS detects N = 1 particle in a given AMP 352 size bin, the composition of the bin is fully characterized. This particle number corresponds to N = 5 353 particles for 4 bins/decade, which is a reasonable number for the particle composition to be reasonably represented by the particles captured, with a resolution over which composition changes may happen in 355 the real atmosphere (Zhang et al., 2004). For simplicity, we scale the fraction of the particles contributing 356 information content linearly for conditions with N < 1. with Zefluor filters (9 cm diameter, 1 mm thick, and 1 μm pore size, from MilliporeSigma Corp., 361 Burlington, MA, USA) with subsequent procedures as described by Dibb et al. (1999Dibb et al. ( , 2000 and impactor (operating at cabin T and ambient P) is expected to size-select dry particles, similar to the AMS. 455 The impactor provides a nominal PM1 cut at T = 293.15 K and P = 1013 mbar but the dta,50 for a given 456 particle is pressure-and temperature-dependent, and thus varies with altitude. For instance, at an aerosol 457 density of 1.7 g cm -3 (the ATom-2 campaign average), dta,air,50 drops from 1 μm to 912 nm at 6 km, and impactor was operated under ambient T (not typically done, and best avoided for an optimal particle cut; 461 summarized in Table S1). Hence, the deviation from the nominal 1 μm cut size can be very significant at high altitude (although it could in principle be modulated by changing the flow rate vs. altitude). The If we compare the AMS transmission to ground-level based dry dta (using a dry particle density 468 of 1.7 g cm -3 to calculate dta from dva), the ATom-2 / 3 / 4 dta,sea,50 are 599 nm, 615 nm, and 758 nm, 469 respectively (the dta,air,50 are higher and listed in Table S1; for example, dta,air,50 is 782 nm and 837 nm at and 10 min averages (Fig. S23). The slightly worse fitting slope of 1.09 and r 2 of 0.93 in ATom-2 may be 523 due to the larger contribution of sea salt in ATom-2 in the boundary layer (Hodzic et al., 2020) and hence 524 the larger uncertainty in applying the AMS size cut. To illustrate the impacts of sea salt, we replotted the 525 comparisons ( Fig. 4a-b) colored by sea salt shown as Fig. S24a-b, which suggests that some outliers in 526 ATom-2 are observed at high sea salt concentrations. We also investigate the potential differences in the 527 data products due to the differences in raw data processing criteria for cloud artifacts between AMS and 528 NOAA size spectrometers and find no clear evidence (Fig. S24c-d). Furthermore, we confirm that 529 excluding submicron dust volume is reasonable; only a few outliers have noticeably higher contributions 530 from dust ( Fig. S24e-f).

531
Species density is used to convert the AMS mass to volume concentrations and thus affects the 532 volume comparison. As discussed above (Fig. S12), in this study is estimated with the 533 parameterization method of Kuwata et al. (2012). The parameterization method from Kuwata et al.,534 (2012) was validated up to 1.9 g cm -3 (i.e., oxalic acid) and the lab generated SOA in that study had up to 535 1.46 g cm -3 with an O/C of 0.72. The estimated ATom-1 and -2 is close to that of succinic acid, 536 1.57 g cm -3 , that has a similar O/C ratio (ATom-1 and -2 vs. succinic acid: 1.05 ± 0.44 vs. 1.0), and falls 537 into the observed density range, 1.5-1.7 g cm -3 , for low mass concentrations of SOA (< 3 μg m -3 , as 538 the most cases in ATom), made from α-pinene and ozone from a chamber study (Shilling et al., 2009 Vchem by subtracting 0.2 g cm -3 from the AMS estimated (Fig. S25). Compared to the base cases ( Fig.   542 4a-b), the r 2 values barely change and the slopes increase by 5% or 8% due to the higher estimated OA 543 volume in Vchem. Therefore, this uncertainty is below 10% and does not undermine the agreement within 544 the uncertainties between Vchem and Vphys,TC.
To illustrate that applying the AMS transmission to Vphys is a prerequisite for a meaningful 546 comparison, Fig. 4c  ATom-1. 556 Box plots, regressions, and correlations were carried out for the separate datasets in each bin of 557 removed Vphys, as shown in Fig. 5a-c. For the combined ATom-1 and -2 data (Fig. 5a), the majority of the 558 volume ratios are distributed around the 1:1 line and within the combined systematic uncertainty range 559 (combined 2σ of AMS and UHSAS, the size spectrometer that overlaps most with the AMS, see Fig. 1).

560
If using the UHSAS data product alone and applying the AMS transmission, the resulting volume is on 561 average 93 ± 9 % in ATom-1 and 87 ± 14 % in ATom-2 compared to Vphys,TC. Therefore, the UHSAS 562 uncertainty is representative of that of Vphys,TC. The Vphys uncertainty depends on particle size range or 563 mode (see Table 1 in Brock et al., (2019)) and the random uncertainty in Vphys is expected to be smoothed 564 out with longer averaging time scales. All five bins show high correlations with r 2 of 0.79-0.96, with a 565 lower correlation at the 80-100% Vphys removal bin. The smallest slope of 0.84 is also seen at this bin, 566 where the largest discrepancy is expected due to the combined sharpness of the decreasing AMS 567 transmission for larger particles and the rising tail of coarse mode particles into the submicron size range 568 (e.g., the AMS transmission excludes on average 89% of the total sea salt volume sampled during ATom-569 2). When investigating ATom-1 and ATom-2 independently, ATom-1 averages are slightly below unity 570 but consistent throughout the five bins (Fig. 5b), and ATom-2 shows an increasing bias above 60% Vphys 571 removal (again likely due to the much higher sea salt fractional contribution for this campaign). Only the 572 80-100% bin in ATom-2 has substantial data outside the 2σ uncertainty range. Overall, the above results

588
The above discussion demonstrates the critical role of well-characterized AMS transmission for 589 meaningful volume intercomparison. In this section, we aim to quantify the impact of the AMS 590 transmission on the volume comparison by artificially adjusting the transmission with a series of 591 sensitivity tests. As shown in Fig. 6a, the AMS transmission can be characterized by four "anchoring" 592 particle sizes, representing 0% and 100% transmissions at both ends. During ATom-1 and -2, these 593 anchoring sizes (in dva) were estimated as (i) 35 nm, (ii) 100 nm, (iii) 482 nm, and (iv) 1175 nm, 594 respectively, as discussed above (Fig. 2). Uncertainty ranges are estimated for the latter two sizes from 595 the ATom calibrations and shown in Fig. 6d-e. We alter one anchoring size at a time, recalculate Vphys,TC, 596 and re-compare to Vchem, which is kept unchanged. The resulting slopes and r 2 are summarized in Fig. 6. 597 The adjustments at the two lower anchoring sizes, up to ± 25 nm at 35 nm and ± 50 nm at 100 nm, have 598 a negligible impact on the volume comparison due to the small volume/mass concentrations at these sizes 599 during ATom (e.g., Fig. 3), except for the unrealistic 50 nm decrease at 100 nm (the second anchoring point). In contrast, a dependency of the fitting results on the details of the AMS transmission curve for 601 large particles is observed. For the third anchoring point, corresponding to the largest particles with 100% 602 transmission (Fig. 6d), a smaller dva excludes more Vphys and results in a higher slope. For example, at the 603 lower one SD limit dva of 445 nm, the fitting slopes increase from 0.97 to 1.01 for ATom-1 and 1.09 to 604 1.12 for ATom-2. These small changes in slope are the largest among the four anchoring points, and they 605 are statistically significant because the changes are one magnitude higher than the fitting 1σ uncertainties respectively, compared to 0.96 and 1.09 derived from applying the ATom-1 and -2 transmission (Fig. 4). standard ground-level PM1 (the most common definition of "submicron") instrument would detect. In this 624 study, we use the standard cut URG cyclone operating at the surface ambient humidity as the reference, 625 simulating its operation at ground sites at different altitudes (e.g., sea level and mountain sites). As 626 discussed above, both the AMS and the AMP size distributions measure dry particles while the "standard" 627 PM1 is defined with practical size-selection under ambient humidity. To account for the difference, the URG transmission is applied to the estimated ambient particle size before losing liquid water content (the 629 effect of water on ρp is also considered) (DeCarlo et al., 2004). We assume no size dependence for ρp or 630 the volume fraction of liquid water content for the submicron aerosols. Ambient P and T from ATom are 631 applied to the URG transmission to account for the shift at non-STP conditions, which is relevant when 632 operating such a cyclone at higher altitudes e.g., a mountain site. The results of applying the AMS and 633 URG PM1 standard cut cyclone transmissions to Vphys are shown in Fig. 7. AMS observed on average 96 634 ± 16% (median 96%) and 94 ± 12% (median 94%) of the volumes that would transmit through a ground-635 level URG PM1 cyclone in ATom-1 and -2, respectively. Although we previously concluded that the 636 AMS was approximately an equivalent ground-level PM0.75 instrument in ATom-1 and -2, the difference 637 in collected volume is only ~5%. This is because the submicron volume size distribution peaked around 638 300 nm (dta; see Fig. 3 for example), where AMS transmission is ~100%, and also due to the effect of 639 liquid water on particle size. 640 Next, we compared the submicron volumes observed from the CU AMS and a MOUDI 1 µm 641 stage impactor during aircraft studies, using the ATom conditions ( Fig. 7c & 7d). The two inlets size-642 select dry particles due to sample line heating. AMS observed 87% and 83% by means, 90% and 85% by 643 medians, in ATom-1 and -2 of that from an airborne MOUDI impactor, lower than the ratios when 644 comparing to the URG PM1 cyclones for two reasons: the smaller cutoff size of URG vs. MOUDI due to 645 particle water and lower operating T for URG (which relates to air viscosity). We also compared the 646 Vphys,TC to the (total) Vphys ( Fig. 7a & 7b). AMS collected 68% by means (the same for ATom-1 and -2, 647 and 78% in ATom-1 and 71% in ATom-2 by medians) of Vphys; in other words, 32% of Vphys was excluded 648 by applying the AMS transmission. For both ATom-1 and -2, there was considerable variability on the 649 fraction of Vphys removed to obtain Vphys,TC, which spanned the range from 0% to 100% removal, thus 650 providing a good scenario of testing the AMS transmission. Nevertheless, this data shows that on average 651 the AMS captured the submicron range well, as shown in Fig. 4, and that the comparisons presented here 652 are meaningful for a wide range of scenarios.
3.6 Characterization of the observable particle populations for different chemical instruments 654 The different parts of the aerosol population included in different measurements and models make 655 comparisons of aerosol species inherently more complex than for gas-phase species. In this section, we 656 characterize the size ranges that contribute information to each composition measurement. Importantly, (median 99.9%), 78% (87%), 68% (74%), and 54% (55%) of Vphys (total AMP particle volume), and 98% 691 (99%), 89% (93%), 41% (41%), and 5% (1%) of the total AMP particle number, respectively. The size 692 range above dp = 100 nm, for which PALMS-AMP (Froyd et al., 2019) reports chemical products 693 (partially by extrapolating composition measurements of others sizes, especially at higher time resolutions 694 and lower concentrations), covers 76% (83%) and 11% (5%) of the AMP volume and number, 695 respectively. 696 To complete the illustration of the coverage of the previously discussed instruments, the vertical 697 profiles of observed volume fractions, in both the submicron range and the full AMP size range, are 698 summarized in Fig. 9 (and the statistics summarized in Table S2  shows the heterogeneity and complementarity between PALMS-AMP and the other submicron bulk measurements as a function of altitude. The differences between the 3 min characterization and the 709 PALMS-AMP products are greatly reduced by averaging to 60 min. 710 In summary, outside dust or biomass burning plumes, the particle volume sampled by AMS is 711 within 97 ± 14% compared to SAGA MC, for which the difference disappears for the higher altitude legs, 712 and 85 ± 10% of an airborne dry PM1 measurement, a MOUDI impactor often used in aircraft. AMS and 713 PALMS particle compositional data overlap for a large part of the volume distribution in ATom, and they 714 complement each other at the ends of the distribution (the statistics of the overlap are listed in Table S2). 715 Last but not the least, SAGA filters characterize the particle bulk chemical components representative of 716 the combined size range from the NOAA particle spectrometers.