3 Results
3.1. Statistics of the OXL:SO42- Ratio
Across multiple environments (Fig. 1), we observe similarities in the range of the OXL:SO42- ratio (i.e., slope) across campaigns except for CAMP2Ex (discussed in Section 3.2). A combined plot of all campaigns suggests a common range of OXL:SO42- across different environments (Fig. S1). Visually, the range in OXL:SO42- (Fig. S1) decreases with increasing concentrations, perhaps from a clearer signal of aqueous processing at higher concentrations.
Bootstrapping across all campaigns reveals a stable median OXL:SO42- of 0.0217 (R = 0.76; N = 2948) (details in Table S1) with 95% confidence interval bounds (0.0154 – 0.0296) indicating a relative uncertainty range of ~±30% about the median. The bootstrapped statistics are supported by averaging the OXL:SO42-slopes between campaigns in Fig. 1 (mean: 0.0184, standard deviation: 0.007, median absolute deviation (MAD): 0.004), which, similar to our bootstrapping method, weights all campaigns equally and removes the statistical bias towards larger datasets. The bootstrapped 95% confidence interval of our observed OXL:SO42- (0.0154 – 0.0296) is in excellent agreement with the ratio of yields between aqueous SOA (0.008 – 0.033; Ervens et al., 2011) and SO42- (approximately unity), corroborating the hypothesis of a general range in OXL:SO42-. Numerous studies spanning a diverse set of environments have reported OXL:SO42- values (Table S2) with several falling within our bootstrapped range.
Cumulative probability functions (CDFs) were plotted to more easily compare OXL:SO42- between campaigns (Fig. S2). As point-by-point ratios are sensitive to background levels of OXL and SO42-, we plotted CDFs of their enhancement ratio (ΔOXL/ΔSO4) where values were subtracted by their 10th percentile to approximate background levels per campaign. The resulting CDFs showed similar curves between multiple campaigns and revealed that approximately 20% of point-by-point OXL:SO42- values fall within our bootstrapped 95% confidence interval (0.0154 – 0.0296) while ~50% of OXL:SO42-values fall within 0.010 – 0.050. In comparison with the linear regression slope, point-to-point ratios are more susceptible to differences in background, resulting in differences in agreement with our 95% confidence interval (e.g., CHECSM). This is a consequence of the point-by-point calculation: although the dataset as a whole may have a mean slope within our confidence interval, there may still be variability in the OXL:SO42- ratios of individual points. Thus, as the 10th percentile merely accounts for campaign backgrounds, the linear regression slopes (Fig. 1) may better handle different environments by implicitly accounting for individual background levels via non-zero intercepts.
Surface OXL:SO42- values from CHECSM agree with the bootstrapped 95% confidence interval for non-BB samples for PM18 (0.0264; Fig. 1g) and PM1(0.0196) modes (R = 0.73 for both). Size-resolved data show that this agreement is greatest between 0.32–1 μm (Fig. S3a), where OXL and SO42- masses mostly reside (Cruz et al., 2019). An increase in supermicrometer OXL:SO42- (Fig. S3a) suggests the enhancement of OXL via gas-particle partitioning of OXL and/or its precursors onto coarse particles as documented for the CHECSM region (Stahl et al., 2020b; Stahl et al., 2021). These results suggest the ratio may be applicable to the mixed layer for submicrometer particles.
AToM provides insight into the OXL:SO42- ratio over remote marine environments in both hemispheres. The Pacific and Atlantic Oceans (< 3 km AGL) have a combined OXL:SO42- ratio of 0.0207 (R = 0.51) (Fig. 1f). Separately, the Pacific and Atlantic have ratios of 0.0180 (R = 0.36) and 0.0251 (R = 0.72), respectively, remarkably similar to other environments (Fig. 1). Across altitudes (Fig. S4), OXL:SO42- values for the Pacific between 0 – 7.5 km AGL are within our 95% confidence interval (Fig. S4a) but only near-surface Atlantic samples fall within our confidence interval (Fig. S4c), possibly due to OXL and/or its precursors undergoing gas-particle partitioning onto Saharan dust.
Variability in OXL:SO42- across campaigns was most evident in MASE-II (Fig. 1d) and CAMP2Ex (Fig. 1h), which had instances of very low OXL:SO42-, attributable to (1) fresher plumes that have not had time to form OXL (Wonaschuetz et al., 2012), (2) the degradation of OXL into CO2 (Zhou et al., 2020), (3) the formation of OXL-metal complexes (Sorooshian et al., 2013; Tao & Murphy, 2019), and (4) the presence of high SO42- backgrounds. Correlation coefficients below 0.50 (Fig. 1) signify the presence of confounding sources, an expected result given the diversity of environments analyzed. Seasonal factors may also influence the ratio (Tao & Murphy, 2019). Variability between campaigns (Fig. 1) may be suggestive of SO42- from cloud-free photochemistry and gas-phase oxidation (Ervens, 2015), which are important sources of uncertainty when using the OXL:SO42-ratio to assess aqueous processing, as our proposed range implies that SO42- is mainly from aqueous processing.
The main utility of this ratio is to gauge relative rather than absolute extents of aqueous processing between air masses via a comparison of inferred OXL. Considering the differences between OXL and SO42- in terms of their precursors, formation mechanisms, and sinks, the consistency of the OXL:SO42- ratio across multiple environments implies a convergence towards a fairly narrow range, which is assisted in part by the large sample sizes used in this study.