Seasonal Differences in Correlations and Contributions of Photochemical, Upwind Cloud and Aerosol Aqueous-Phase Oxidation, and Mixed Combustion in Secondary Organic Aerosol Formation in Coastal Southern California

Berta, Veronica Z.; Russell, Lynn Monica; Williams, Abigail S.; Dedrick, Jeramy L.; Han, Sanghee; Lubin, Dan; Farley, Ryan N.; Aiken, Allison C.; Wentzell, Jeremy; Liggio, John

doi:10.5194/egusphere-2026-4695

Preprints

https://doi.org/10.5194/egusphere-2026-4695

Preprints

24 Aug 2026

| 24 Aug 2026

Seasonal Differences in Correlations and Contributions of Photochemical, Upwind Cloud and Aerosol Aqueous-Phase Oxidation, and Mixed Combustion in Secondary Organic Aerosol Formation in Coastal Southern California

Veronica Z. Berta, Lynn Monica Russell, Abigail S. Williams, Jeramy L. Dedrick, Sanghee Han, Dan Lubin, Ryan N. Farley, Allison C. Aiken, Jeremy Wentzell, and John Liggio

Abstract. Organic aerosol (OA) formation in coastal environments is influenced by photochemical and aqueous-phase oxidation, but their relative contributions remain poorly constrained. During the Eastern Pacific Cloud Aerosol Precipitation Experiment (EPCAPE), Aerosol Mass Spectrometer measurements at Mt. Soledad in La Jolla, California were analyzed by positive matrix factorization to resolve four oxygenated OA (OOA) factors: sulfate-related (SR-OOA), more-oxidized (MO-OOA), less-oxidized (LO-OOA), and continental (C-OOA). SR-OOA was linked to marine biogenic sources and bimodal number distributions indicating in-cloud aqueous reactions. Multiple linear regression (MLR) associated SR-OOA with in-cloud aqueous reactions represented by upwind cloud vertical fraction (UCVF; 70 %). MO-OOA correlated with UCVF during months with increasing UCVF (R = 0.25–0.66) and ozone for 11 months (R = 0.36–0.76), indicating both in-cloud aqueous and photochemical oxidation. MLR showed contributions from in-cloud aqueous reactions in spring (48 %) and photochemical oxidation represented by ozone in summer, fall, and winter (44–77 %) to MO-OOA. LO-OOA showed an ozone correlation (R = 0.37) and midday maxima, while MLR associated LO-OOA with photochemical oxidation (64 %). C-OOA correlated with refractory black carbon (rBC; R = 0.33) and other combustion tracers. MLR associated C-OOA with combustion represented by rBC in winter (49 %), aerosol water aqueous oxidation represented by relative humidity (RH) in spring (37 %), and photochemical oxidation in summer and fall (37–43 %). O/C was explained by RH (39 %) and ozone (38 %), followed by UCVF (19 %). These results reveal distinct seasonal contributions of photochemical and aqueous-phase oxidation to biogenic and mixed combustion OA in coastal Southern California.

Received: 04 Aug 2026 – Discussion started: 24 Aug 2026

Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Chemistry and Physics.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2254 KB)

Supplement (1873 KB)

Download & links

Veronica Z. Berta, Lynn Monica Russell, Abigail S. Williams, Jeramy L. Dedrick, Sanghee Han, Dan Lubin, Ryan N. Farley, Allison C. Aiken, Jeremy Wentzell, and John Liggio

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-4695', Anonymous Referee #1, 14 Sep 2026
GENERAL COMMENTS
This study aims to quantify the contribution that different mechanisms have on the formation of secondary organic aerosol (SOA), i.e., formation of SOA in the gas phase, in aqueous aerosol, and in cloud droplets. To address this question, the authors analyze data collected during the year-long EPCAPE field campaign in Southern California, USA. EPCAPE collected measurements of aerosol chemical composition and size distribution, gas chemical composition, meteorological and cloud-related parameters, and back-trajectory of air parcels. These measurements are subsequently analyzed using positive matrix factorization (PMF) to determine source apportionment, and multilinear regression (MLR) is used to determine the correlations and contributions of different variables to SOA formation. Sub-setting the data by season and air mass back-trajectory allows for identifying the patterns in season and regional influence, respectively.
The reviewer considers that the methodology used in this study to collect and analyze the data is rigorous. Furthermore, the reviewer considers that this study addresses an important and relevant question in the atmospheric sciences, and publishing this study in Atmospheric Chemistry and Physics would contribute to the atmospheric science research community.
The reviewer's primary critique of this study concerns the multivariate linear regression, which is described in the Specific Comments section.

SPECIFIC COMMENTS
Critiques on multivariate linear regression:
Why are the predictor variables normalized? What is physical and mathematical reasoning behind it?

If the predictor variables (X₁, X₂, etc.) are normalized, shouldn’t the predicted variable (Y) also be normalized?

In order to stay consistent with Equation 1, and to further emphasize that the multivariate linear regression is being fed normalized predictor values, shouldn’t the predictor variable in Equation 2 have the subscript “norm”, i.e., be “X_i,norm” instead of just “X_i”? Also, for consistent notation, in Line 1088, the average should read “X_i,norm”, not “X_norm,i”.

Line 208 and 209. As a consequence of the previous comment, the subscripts in the text should include “norm”, i.e. they should read “… normalized predictor value (X_p,norm)…” and “… all m_pX_p,norm predictor terms …”.

The reviewer understand that the researchers have the freedom to make assumptions in how to best model a multivariate regression. However, could you please provide some reasoning as to why the intercept (b) in the regression is set to zero? And if it is not set to zero, does it change the final conclusions of this study?

Line 498. The correlations in Figure 7f are described as “strong” with R values of 0.43 and 0.49. However, Line 111 defines a strong correlation as one with R ≥ 0.8. Please ensure that the terminology used to describe correlation strength is consistent throughout the manuscript.

Other critiques:
Line 18. Perhaps it is better to use the phrase “cloud processes” instead of “in-cloud reactions” because the bimodal nature of the size distribution is not due to chemical reactions alone; it is also due to the process of scavenging, which is a physical, not chemical, process (Dedrick et al., 2026).

Line 50. The phrase “photochemical oxidation” is used at the beginning of the sentence and then “photochemical aging” is used at the end. Are the authors using these two phrases interchangeably? If so, please consider using consistent terminology.

Line 162. Line 110 says that CVI measurements were excluded from this study. Please clarify.

Figure 2. The three-month periods used to calculate the seasonal averages differ slightly from the traditional definitions of meteorological seasons (NOAA). The reviewer recognizes that this difference is unlikely to affect the study’s findings; however, please explain the rationale for using these alternative three-month periods.

Figure 7a–e. Because the correlation coefficient is used in the text to characterize the relationships shown in these subplots, the reviewer suggests including the corresponding R (or R²) values in Figures 7a–e for clarity.

Figure 7 caption. Since measured O/C is on the x-axis, the correct description for Figure 7f should be “O3 vs O/C”, rather than “O/C vs O3”. The same correction applies to Figure 7g–j.

Figure 7a and 7f. The time scales used in subplot (f) are annual, fall, and spring time scales. However, the time scales for the other subplots (g)–(j) are not specified. For clarity, could you please specify the time scale used in each subplot?

Figure 8. Could the authors please clarify the purpose of the red polynomial fits shown in subplots (k)–(t) and explain why they are not discussed in the text? The reviewer suggests either removing these fits if they are not essential to the analysis or providing a brief explanation of their purpose and interpretation in the text.

TECHNICAL COMMENTS
Line 111. Are the definitions of “weak”, “moderate”, and “strong” in terms of R or the absolute value of R (|R|)? In Line 578, the absolute value notation is used.

Line 131. There should be a dash in “high-resolution”.

Line 132. The negative sign in “I^-” must be superscripted.

Lines 188, 506, 507. Please use the en dash for range of numbers, and the dash for negative numbers.

The in-text citations must have a space after the semicolon.

The bibliography should use a consistent formatting. For instance, currently some references capitalize the name of the journal while others don’t.

REFERENCES
Dedrick, J. L., Osawa, A., Russell, L. M., Miller, M. A., Robinson, L., Chang, R., Williams, A. S., Berta, V. Z., Ravichandran, E., Petters, M., Wheeler, M. J., Han, S., Pelayo, C. N., Maneenoi, N., Lubin, D., Uin, J., Enekwizu, O., Zawadowicz, M., Mayol-Bracero, O., & Twohy, C. H. (2026). Quantitative Estimates of Upwind and On-Site Particle Scavenging and Retrieved Cloud Supersaturations from Urban Coastal Stratocumulus during EPCAPE. Journal of the Atmospheric Sciences, 83(6), 1061–1082. https://doi.org/10.1175/JAS-D-26-0004.1

National Oceanic and Atmospheric Administration. (2024, February 27). Infographic: Meteorological and astronomical seasons. https://www.noaa.gov/education/multimedia/infographic/infographic-meteorological-and-astronomical-seasons
Citation: https://doi.org/10.5194/egusphere-2026-4695-RC1
RC2:
'Comment on egusphere-2026-4695', Anonymous Referee #2, 22 Sep 2026
Based on observations from the Eastern Pacific Cloud Aerosol Precipitation Experiment (EPCAPE), this study investigated the seasonal contributions of photochemical and aqueous-phase oxidation to biogenic and mixed combustion organic aerosol (OA) in coastal Southern California. These findings are valuable for elucidating the relative roles of these oxidative processes in OA formation within coastal environments. Here I have some suggestions for authors to further clarify their results as the below:
Given that SR-OOA is primarily driven by marine biogenic sources and daytime sea-breezes transporting marine-influenced air masses inland, why does its diurnal cyclepeak in the evening? (Fig. 2a–e)?

Why does the minimum of O₃ appear in summer( August), rather than in winter? (Fig.3a). This is contradictory to the daily mean downwelling shortwave radiation (DSW) peaking during summer.

The organic and inorganic components had very low diurnal variability across seasons due to the daily land-sea breeze transitions characteristic of the coastal environment as well as frequent daytime cloudiness reducing the role of photochemically driven changes (Han et al., 2025) (L352-355). This sentence is very abrupt. I can’t follow the logic between it and the surrounding sentences.

Peak daily mean downwelling shortwave radiation (DSW) occurred during summer (235 ± 324 W m^-2, Table S7), but LO-OOA concentrations remained low, suggesting a reduction of photochemical efficiency by the increased local cloud fraction in May (0.69 ± 0.39, Fig. 3d) (L363-366). I can't agree with this explain. The photochemical efficiency is closely related to DSW.

The current organization of Section 3.2could be improvedto enhance logical flow.

Section 3.3, a figure showing the back-trajectories is recommended, which helps to distinguish different air masses.

It is hard to understand that "Ozone has a stronger contribution to variability during winter (58%) and fall (70%), and a smaller role during spring (15%) and summer (30%)" (L504-505). Same as Q2.

The conclusionsshould be refinedto better highlight the key findings.
Citation: https://doi.org/10.5194/egusphere-2026-4695-RC2
RC3:
'Comment on egusphere-2026-4695', Anonymous Referee #3, 30 Sep 2026
This manuscript presents a useful year-long dataset on organic aerosol composition at a coastal site in Southern California. The combined analysis of PMF factors, chemical tracers, and upwind cloud conditions is interesting and relevant to ACP. However, I have concerns about how the size distributions and regression results are interpreted. I recommend major revision.
Major comments
The size distributions in Fig. 1 need clearer explanation. From the caption, these are total particle number distributions during periods when one PMF factor exceeds its 80th percentile, rather than size distributions of the individual factors. Please make this distinction explicit in the Methods and discussion. The authors should report the number of observations in each group and check whether the differences persist within comparable seasons and air masses. A Hoppel gap suggests cloud processing, but does not necessarily show that the associated organic factor was formed in cloud water.

My main concern is the interpretation of the MLR percentages as contributions from different formation pathways. The calculation based on regression coefficients multiplied by mean normalized predictors is not a partitioning of explained variance or SOA production. Thus, statements such as “70% of SR-OOA formation” being attributed to upwind cloud reactions appear too strong. Please clarify what these percentages represent and revise the Abstract, Conclusions, and corresponding figures accordingly.

The regression setup also needs justification. Min–max normalization does not resolve non-normality, and linear regression does not require normally distributed predictors. Why was the intercept fixed at zero? After normalization, zero represents the observed minimum rather than the absence of the relevant process. Please compare the results with a model that includes an intercept. Table 1 reports uncentered R2 which should not be interpreted as the fraction of variability explained. Conventional R2 and an independent validation would provide a more informative assessment.

Have the authors checked whether the linear model adequately describes these relationships? Nonlinear responses and interactions are plausible, particularly for RH and cloud-related variables. A comparison with a GAM or a machine-learning model, such as random forest, would help assess the robustness of the results. I do not consider machine learning essential, but the choice of MLR should be supported by diagnostics and validation. Any alternative model should be evaluated using time-based holdouts, and its variable importance should not be treated as a chemical formation fraction.

The process proxies require more cautious interpretation. RH is not equivalent to aerosol liquid water, and UCVF does not directly measure the cloud exposure of the sampled particles. Both may reflect changes in air-mass origin, precursor availability, or transport. Could aerosol liquid water be estimated from the available measurements? The authors should also discuss how excluding local in-cloud observations affects their comparison of local and upwind cloud influences.

Please provide more information on PMF stability, including sensitivity to initialization and rotation, and whether the factor profiles remain representative across seasons. The increase in Q/Qmathrm from the two-factor to the three-factor solution in Table S2 also needs explanation. The wildfire sensitivity analysis is useful, but the resulting changes in regression contributions should be more clearly acknowledged when presenting seasonal conclusions.

Minor comments
7. Please check the subtraction direction used to define “OminusC” in line 170.
8. The C-OOA N/C value in the text differs from Table S3, and some seasonal values appear inconsistent with Table S7.
9. Please check the duplicated captions of Figs. S6 and S9, outdated table references in Text S1, and the “Table S#” placeholder in Fig. S10.
10. UCVF should be defined consistently as either “vertical fraction” or “volume fraction.”
11. Please explain the weighting used to calculate O/C in Fig. S14 and report how much data were removed by the 8000 cm-3
Citation: https://doi.org/10.5194/egusphere-2026-4695-RC3

Veronica Z. Berta, Lynn Monica Russell, Abigail S. Williams, Jeramy L. Dedrick, Sanghee Han, Dan Lubin, Ryan N. Farley, Allison C. Aiken, Jeremy Wentzell, and John Liggio

Supplement

https://doi.org/10.5194/egusphere-2026-4695-supplement

Veronica Z. Berta, Lynn Monica Russell, Abigail S. Williams, Jeramy L. Dedrick, Sanghee Han, Dan Lubin, Ryan N. Farley, Allison C. Aiken, Jeremy Wentzell, and John Liggio

Viewed

Total article views: 394 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
221	113	60	394	80	63	73

HTML: 221
PDF: 113
XML: 60
Total: 394
Supplement: 80
BibTeX: 63
EndNote: 73

Views and downloads (calculated since 24 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	77	24	6	107
Sep 2026	118	76	49	243
Oct 2026	26	13	5	44

Cumulative views and downloads (calculated since 24 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	77	24	6	107
Sep 2026	118	76	49	243
Oct 2026	26	13	5	44

Viewed (geographical distribution)

Total article views: 381 (including HTML, PDF, and XML) Thereof 381 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 08 Oct 2026

Short summary

This study examines different formation pathways for various types of organic aerosol using one year of aerosol mass spectrometry measurements from the Eastern Pacific Cloud Aerosol Precipitation Experiment (EPCAPE) in La Jolla, California. The results show that photochemical, cloud and aerosol water aqueous reactions, and combustion emissions contribute differently to organic aerosol across seasons, with in-cloud aqueous reactions most strongly associated with highly oxidized organic aerosol.


Total:	0
HTML:	0
PDF:	0
XML:	0