Plain Language Summary
The extent of atmospheric chemical processing remains an uncertain aspect of air mass characterization. Addressing this uncertainty is important because chemical reactions in the atmosphere in the presence of water (aqueous processing) produce a large fraction of global aerosol mass. The oxalate-to-sulfate ratio has been proposed as an indicator of aqueous processing, where higher values point to increased processing of an air mass. In this study, we quantify a range in the oxalate-to-sulfate mass ratio (0.0154 – 0.0296) using data from multiple field campaigns encompassing a diverse set of environments. This range is robust near the surface for particles below 1 micrometer in diameter. Larger particles, especially dust, and biomass burning particles significantly affect the oxalate-to-sulfate ratio and thus may confound the interpretation of a high oxalate-to-sulfate ratio as a signal of aqueous processing. In the absence of dust and biomass burning particles, the oxalate-to-sulfate ratio range may be used to compare the relative extent of aqueous processing between different air masses.
1 Introduction
Organic aerosol (OA) account for a major fraction of atmospheric aerosol particles (Hallquist et al., 2009; Kanakidou et al., 2005; Q. Zhang et al., 2007) comprising between 20 – 50% of fine aerosol mass in the continental mid-latitudes (Putaud et al., 2004; Saxena & Hildemann, 1996), 30–80% in the free troposphere (Murphy et al., 2006), and over 80% in tropical forests (Andreae & Crutzen, 1997; Roberts et al., 2001). Secondary organic aerosol (SOA) are derived from gas-to-particle conversion processes, including aqueous processing (Blando & Turpin, 2000; Ervens et al., 2011; Warneck, 2003), wherein oxidized volatile organic compounds (VOCs) partition into cloud droplets or wet aerosol particles and undergo chemical reactions to form low-volatility products that remain in the condensed phase (Ervens, 2015; Ervens et al., 2011; McNeill, 2015). The formation of SOA through aqueous processing has been estimated to serve as a global SOA source comparable to the gas-phase pathway, with 90% of aqueous-phase SOA formed in-cloud (Fu et al., 2008). Using the scheme from Fu et al. (2008), Heald et al. (2011) found contributions of SOA to total OA ranging between 20 – 80% across environments. Oxygenated organic species comprise 60 – 95% of total organic aerosol mass across urban and remote sites (Q. Zhang et al., 2007), while SOA from VOCs explains up to 70% of global organic carbon mass (Hallquist et al., 2009). However, despite improvements in modeling organic aerosol (Heald et al., 2005, 2011), atmospheric chemistry models still underestimate SOA (Schroder et al., 2018) due partly to an incomplete understanding and representation of aqueous processes, resulting in poor model parameterization (Hallquist et al., 2009; McNeill, 2015). The inclusion of SOA formation has been shown to decrease model bias (–64% to –15%) and increase model correlations with observations (R = 0.5 to 0.6) (Carlton et al., 2008); thus, improving SOA estimates is a key area of development for models to more accurately evaluate the impacts of atmospheric aerosol particles and reduce uncertainties regarding the net effect of aerosol particles on health and climate (IPCC, 2014).
Oxalic acid is the most abundant organic acid in tropospheric aerosol particles across different regions (Cruz et al., 2019; Yang et al., 2014; Ziemba et al., 2011). The dissociated ion oxalate (OXL) is a well-established tracer of aqueous processing, contributing 1 – 10 % of total particulate mass (Ervens, 2015; Myriokefalitakis et al., 2011) and 3 – 4% of total organic mass over marine/continental areas (Sorooshian et al., 2010). OXL is often used in combination with other secondary tracers such as SO42- to assess the extent of aqueous processing in a region (Crahan et al., 2004; Hilario, Cruz, Bañaga, et al., 2020; Sorooshian, Varutbangkul, et al., 2006; G. Wang et al., 2012; Yu et al., 2005). Direct sources are thought to be minor relative to production via aqueous processing (Ervens, 2015; Huang & Yu, 2007; Myriokefalitakis et al., 2011) and photochemistry (C. Zhang et al., 2020). Sources of gaseous OXL precursors include biomass burning (BB) (Narukawa et al., 1999; Yang et al., 2014) and biogenic emissions (Boone et al., 2015). Early model simulations overestimated OXL by an order of magnitude (Crahan et al., 2004) while global simulations by Myriokefalitakis et al. (2011) showed better agreement over marine/rural environments between observed and modeled OXL (modeled:observed slope = 1.16 ± 0.14; R = 0.60) but could not capture OXL over urban regions (weak correlation; r ≈ 0). More recent work reports reduced discrepancy between models and observations for SOA concentrations (Hodzic et al., 2016; Pai et al., 2020; Zhu & Penner, 2019).
The OXL:SO42- ratio has been suggested in past work to be an indicator of aqueous processing (Ervens et al., 2014; Wonaschuetz et al., 2012; Yu et al., 2005). The usage of the ratio as an aqueous processing marker implies that OXL and SO42- are entirely sourced from aqueous-phase oxidation, whether it be in cloud droplets or wet aerosol particles, and does not account for gas-phase oxidation in cloud-free air (Ervens, 2015; D. D. Huang et al., 2020; Zhan et al., 2021). This is a good assumption for OXL as there is thought to be no gas-phase reaction that would produce OXL (i.e., an aqueous medium is required for OXL production) (Warneck, 2003); however, OXL is influenced by gas-particle partitioning equilibrium and can exist in the gas-phase as oxalic acid (Nah et al., 2018; Tao & Murphy, 2019). For SO42-, gas-phase oxidation is an important source of uncertainty as it can dominate over aqueous processing at times (D. D. Huang et al., 2020). Though we also note that oxidation in the gas-phase is much slower than in the aqueous-phase (Cautenet & Lefeivre, 1994) and aqueous-phase oxidation explains 60 – 90% of SO42- in global models (Barth et al., 2000; Faloona, 2009; Fu et al., 2008).
The OXL:SO42- ratio can serve as an aqueous processing marker because aqueous media (including clouds and wet aerosol particles) facilitate the production of both OXL and SO42- at rates dependent on liquid water content for SO42- formation and droplet surface area for OXL (Ervens et al., 2011). These two cloud parameters correlate within growing clouds (Kim et al., 2003), which connects in-cloud OXL and SO42-production such that their ratio may lie within some range. The OXL:SO42- ratio has been observed to correlate well with cloud fraction and fall within 0.01 – 0.03 between 0 – 4 km above ground level (AGL) when cloud fractions are high (Wonaschuetz et al., 2012). Therefore, identifying a range in the OXL:SO42- ratio across different environments can be useful for comparing relative extents of aqueous processing with higher ratios suggesting more processed air. This comes with the assumptions outlined above that SO42- is mainly sourced from aqueous processing, which may not hold for certain environments. However, this ratio is expected to be particularly applicable near clouds. It is important to note that this ratio likely exhibits a seasonal cycle as observed in Tao & Murphy (2019). Work by Yao et al. (2004) also demonstrated seasonal shifts in the relative contributions of primary and secondary OXL.
Laboratory experiments are often relied on for mechanistic details of aerosol particles (e.g., Hennigan et al., 2010; Pang et al., 2019) but sometimes disagree with aircraft measurements (May et al., 2014). Thus, aircraft campaigns provide a valuable opportunity to study aerosol particles influenced by cloud processes in their most natural environment (Sorooshian et al., 2020). This study leverages composition data from multiple field campaigns, predominantly based on airborne measurements, to investigate the following questions: (1) Is there a generally consistent range of OXL:SO42- across different regions?; (2) does this ratio depend on particle size?; and (3) what conditions can significantly affect OXL:SO42-values?
2 Methods
This work relies mostly on airborne field datasets, with the focus predominantly being on particle-into-liquid sampler (PILS) data. The PILS converts sampled aerosol particles into droplets sufficiently large to be collected via inertial impaction, with the resultant liquid transported to vials on a rotating carousel for post-collection chemical analysis via ion chromatography (IC) (Sorooshian, Brechtel, et al., 2006; Weber et al., 2001).
Table 1 summarizes relevant details across campaigns, namely: the International Consortium for Atmospheric Research on Transport and Transformation (ICARTT), the Marine Stratus/Stratocumulus Experiment (MASE-I), the Gulf of Mexico Atmospheric Composition and Climate Study (GoMACCS), the Marine Stratus/Stratocumulus Experiment II (MASE-II), the Nucleation in California Experiment (NiCE), the Atmospheric Tomography Mission (AToM), the Cloud, Aerosol, and Monsoon Processes-Philippines Experiment (CAMP2Ex), the ground-based CAMP2Ex weatHEr and CompoSition Monitoring (CHECSM) study, and the Aerosol Cloud meTeorology Interactions oVer the western ATlantic Experiment (ACTIVATE). Note that species from AToM were collected by Soluble Acidic Gases and Aerosol (SAGA) filters (Dibb et al., 2002, 2003). We use ground-based data from Metro Manila, Philippines during CHECSM collected by a micro-orifice uniform deposit impactor (MOUDI) and analyzed via IC to assess how the OXL:SO42- ratio may vary within the mixed layer and across particle sizes. We examined a total of 53 MOUDI sets that were collected on a weekly basis with a sample duration of ~48 hours per set. Six of those sets were impacted by BB based on the criteria presented by Gonzalez et al. (2021) for the same dataset.
As we analyze a CAMP2Ex case study in Section 3.2, we summarize the campaign here with more details provided in Hilario et al. (2021). CAMP2Ex took place over the Western Pacific and aimed to study the influence of radiation, convection, and meteorology on aerosol and gas species. With 19 research flights between August to October 2019, CAMP2Ex provided a rich composition, radiation, and convection dataset spanning 0 – 9 km AGL. In this study, we analyze submicrometer non-refractory aerosol from the aerosol mass spectrometer (AMS; Aerodyne) (Canagaratna et al., 2007; DeCarlo et al., 2006) and size distributions collected by a laser aerosol spectrometer (LAS; TSI Model 3340) for particle diameters 50 – 3162 nm. A comparison of SO42- from the PILS and AMS during CAMP2Ex shows good agreement (AMS:PILS slope = 0.81; R = 0.88), suggesting that SO42- was predominantly in the submicrometer size range given the size ranges of the AMS (PM1) and PILS (PM~4). Based on sea salt SO42- calculations, less than 5% of SO42- during CAMP2Ex originated from sea salt.
During AToM, CAMP2Ex, and ACTIVATE, the aerosol sampling inlet likely limited the upper size to approximately 4 μm (McNaughton et al., 2007) although there may have been additional impaction losses in the sampling lines internal to the aircraft that further smoothed the particle transmission curve near this upper bound. This higher cutoff size allowed for sampling of sea salt and dust.
To meaningfully compare OXL:SO42-ratios between campaigns, we separated out samples impacted by strong point sources (e.g., ship plumes, cattle feedlots, smoke) as identified using flight scientist notes and clear enhancements in particle concentration data. When calculating species ratios, we excluded instances when the denominator value was below its 5thpercentile to reduce the uncertainty caused by low denominator values. Ratios of species in this study refer to mass ratios unless otherwise indicated (e.g., molar ratios). To quantify an all-campaign statistic and uncertainty, median and 95% confidence intervals of the OXL:SO42- ratio were derived via bootstrapping of all campaigns using different combinations of sample size and number of iterations, excluding samples with confounding influence (further details are provided in Table S1).
Table 1 . Details of aircraft campaigns analyzed. Statistics of oxalate (OXL) and sulfate (SO42-) collected by PILS from Figure 1 are listed (R = Pearson correlation, N = number of samples). Median absolute deviation (MAD) is provided as a measure of slope uncertainty (e.g., slope ± MAD; r, N). Statistics are presented separately for biomass burning (BB) and non-biomass burning (non-BB) samples.