the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 May 2026
Marine Organic Aerosols Reflect Ecosystem Variability from Phytoplankton Functional Types to Micronekton
Abstract. Marine organic aerosols remain a major source of uncertainty in aerosol cloud–climate interactions, in part because marine ecosystem structure and biological drivers are often represented in overly simplified terms, typically reduced to bulk chlorophyll‑a. Here, a full year of high-resolution aerosol mass spectrometry measurements at Mace Head (west coast of Ireland) is combined with HYSPLIT air-masses exposure metrics and gap-free phytoplankton functional type (PFT) fields to explore influences on primary marine organic aerosol (PMOA) and methane sulphonic acid (MSA). During the spring-summer diatom climax, PMOA correlates with dominant bloom taxa (R=0.65-0.70) and micronekton (R=0.55), with rapid 1-3 day responses and secondary maxima at ~25 days, consistent with early labile release and later lysis/grazing. During that same phase, MSA also showed a lagged responses to both PFT and micronekton reflecting delayed DMS production and oxidation. However, comparable phytoplankton air-mass exposure in the late summer of that same year (i.e. early depletion phase) did not reproduce such high correlations, with time-scale analyses indicating weakened coupling at warmer sea-surface temperatures despite moderately stronger winds. These results imply that structured ecosystem composition and physical forcing both contribute to cross-basin seasonal differences in marine organic aerosols formation. This motivates future research vessel campaigns and mesocosm experiments to explicitly manipulate PFT interactions and air-sea physics.
- Preprint
(1777 KB) - Metadata XML
-
Supplement
(3455 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-2745', Anonymous Referee #1, 18 Jun 2026
-
AC2: 'Reply on RC1', Emmanuel Chevassus, 20 Aug 2026
General Assessment
This manuscript presents a comprehensive year-long study combining high-resolution aerosol mass spectrometry with gap-free phytoplankton functional type (PFT) data and HYSPLIT trajectory analysis at the Mace Head station. The work demonstrates that PFT-specific linkages to PMOA and MSA are more nuanced than bulk chlorophyll-a approaches, with distinct lag structures and seasonal shifts in the dominant PFT-aerosol couplings. The methodology is sophisticated, combining PMF source apportionment, cross-entropy lag analysis, and wavelet coherence.
The study makes a valuable contribution to understanding marine aerosol sources, but several methodological and interpretive aspects require clarification before publication.
The following issues should be clarified to improve the manuscript further:
We thank Reviewer 1 for the careful reading and thorough constructive comments. Each point is addressed below (author responses in italic).
- PMF Source Apportionment
Q1a: The PMOA factor was constrained using reference spectra from Ovadnevaite et al. (2011a) with a-values ranging from 0.1 to 0.5. How sensitive are your PMOA concentrations to the choice of a-value? You mention that 3-5 factor solutions showed unrealistic behavior, but what about varying a-values within the 6-factor solution? A sensitivity analysis should be presented or referenced in the supplement.
Response: A sensitivity analysis of PMOA to the a-value within the retained six-factor solution is presented in Figure S9, where the a-value is varied from 0.1 to 0.5 in steps of 0.05; the reconstructed PMOA time series and its correlations with external tracers (eBC, PFTs) are stable across this range, and a = 0.40 was retained. This is complemented by the 300-run bootstrap of Figure S10. We have added an explicit cross-reference to Figure S9 in Section 3.2 so this sensitivity is easier to locate.
Q1b: The PMOA mass spectrum shows 55% oxygenated carbons and 39% aliphatic fragments. How does this fingerprint compare to the reference spectrum you used for constraint? Is there evidence of significant deviation that might indicate mixing with other sources?
Response: Because the PMOA factor is constrained to the Ovadnevaite et al. (2011a) reference spectrum, the a-value directly bounds the permitted deviation from that fingerprint; the retained solution (a = 0.40, Figure S7A) reproduces the characteristic 55 % oxygenated / 39 % aliphatic signature with the diagnostic CxH2y-3, CxH2y-1CO and CxH2y+1 ion series, consistent with the established Mace Head PMOA profile (Ovadnevaite et al., 2011b; Chevassus et al., 2025). The main evidence against mixing with other (e.g. combustion) sources is the low correlation with eBC (R = 0.16, Figure S9), compared with the inflated values in the under-fitted 3-5 factor solutions (R = 0.35-0.62, Figure S8). We now state this explicitly in the revised Section 3.2.
Q1c: You excluded pollution events using MAAP/eBC data. What was the threshold for flagging pollution events, and what percentage of data were excluded? This is critical for ensuring the "clean marine" signature you claim.
Response: We should first correct a misunderstanding: we do not apply a hard pollution or clean-sector filter to the dataset, so there is no eBC threshold that removes a fixed fraction of points, and the current Methods wording ('used to flag pollution events') was reworded to avoid implying otherwise. Anthropogenic influence is instead handled intrinsically by the constrained source apportionment: PMF resolves local combustion emissions as their own factor(s), so anthropogenic mass is apportioned away from PMOA. eBC is retained as an independent counterfactual, and the low PMOA-eBC correlation (R = 0.16, Figure S9) confirms that the PMOA factor is not driven by pollution episodes. This mirrors our prior source apportionment at the same site with the same instrument (Chevassus et al., 2025), where the resolved PMOA factor also showed no eBC correlation (R = 0.17). We have reworded Section 2.2 accordingly. Both references are already cited in the manuscript.
- PFT Data and Exposure
Q2a: The AIGD-PFT dataset has 4 km resolution, but your HYSPLIT trajectories use a 20 km radius for exposure calculations. How do you account for the spatial scale mismatch between PFT patchiness (kilometer-scale features noted on line 252) and trajectory averaging?
Response: The scale mismatch is handled in two ways. First, exposure is computed as the geometric mean of surface concentration within the sampling radius (not a linear average), which is the appropriate estimator because surface biological activity is approximately log-normally distributed and is therefore dominated by patchy high-concentration features. Second, we explicitly tested radius sensitivity from 20 to 80 km (Figure S2): because fronts and eddies advect at 50-80 km per day, these radii span the relevant patch scales, and the resulting exposure time series were consistent, indicating limited sensitivity to buffer size. The native 4 km PFT field is thus retained and averaged only over the transport-relevant footprint. This clarification is now stated explicitly in Sect. 2.5, with the radius-sensitivity analysis shown in Figure S2.
Q2b: You filtered trajectories with endpoints below 850 hPa and boundary layer heights below 50 m. What fraction of total trajectories were excluded by these filters?
Response: We would clarify the filtering, as the two criteria act differently. Trajectory endpoints with pressure below 850 hPa were assigned zero exposure (decoupled from the surface), whereas boundary-layer heights below 50 m were not discarded but floored to 50 m to avoid unphysical values (Yan et al., 2024). The 850 hPa criterion sets ~5 % of endpoints to zero. We have added this fraction and clarified the distinction in Section 2.5. Gaps were imputed using Kalman smoothing because the exposure fields constitute regularly sampled, serially dependent time series, for which state-space methods estimate missing observations from the temporal structure of the series instead of replacing them with fixed summary values or deleting observations. This is the statistical basis of the na_kalman implementation in imputeTS (Moritz and Bartz-Beielstein, 2017). which was specifically developed for univariate time-series imputation. Comparative evaluations have likewise identified Kalman-based approaches among the most effective methods for univariate time-series imputation (Junger and Ponce De Leon, 2015), including environmental applications in which structural time-series Kalman smoothing outperformed mean, spline and moving-average alternatives for randomly missing air-quality observations (Wijesekara and Liyanage, 2020).
Q2c: The exposure calculation assumes aerosols are well-mixed within the boundary layer. Did you test the sensitivity to using a different vertical mixing scheme or decay rate for organic aerosols during transport? This is particularly relevant for the 72-hour back trajectories, where loss processes could be significant.
The well-mixed boundary-layer assumption with loss scaling to transport time is the standard formulation for Lagrangian surface-exposure metrics (Yan et al., 2024). The effect of transport-time-dependent loss is bounded by the trajectory-lifetime sensitivity test (Figure S3): varying the back-trajectory duration from 24 to 72 h left the exposure signal broadly unchanged (Pearson R > 0.7 against the 72-h reference). Because primary organics are largely non-volatile and the dominant lag signals we interpret occur within 1-3 days (Section 3.4), decay over 72 h is unlikely to alter the lag structure. We now state this limitation explicitly and note that an explicit organic decay scheme is beyond the scope of the exposure metric.
- Cross-Entropy and Wavelet Analysis
Q3a: Cross-entropy was used instead of conventional cross-correlation to avoid "spurious lag structure under non-stationary conditions." Did you test both methods and compare results? Figure S4 shows the PMOA-eBC comparison failing cross-entropy, but this doesn't demonstrate that cross-entropy is superior—only that it's more conservative. A direct comparison table would strengthen your methodological choice.
Response: We thank the reviewer for the suggestion, but we believe that a side-by-side table would compare a valid estimator against a mis-specified one as outlined in the next paragraph below. We have now, however, added complementary time-series diagnostics to explicitly evaluate the statistical properties motivating the use of cross-entropy and cross-wavelet coherence. Stationarity was assessed using the augmented Dickey–Fuller (ADF) and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) tests, which evaluate unit-root and stationarity null hypotheses, respectively. The Priestley-Subba Rao (PSR) test was additionally used to assess temporal changes in second-order spectral structure (rejection of spectral stationarity directly motivates using cross-wavelet coherence) which justifies using biwalets.
By "cross-entropy" we mean the entropy cross-dependence function Sρ (Giannerini and Goracci, 2023), a normalised Bhattacharyya-Hellinger-Matusita distance between the joint density f(X_t, Y_{t+k}) and the product of marginals f(X_t)·f(Y_{t+k}). It equals 0 iff X_t and Y_{t+k} are independent, is bounded in [0,1], and unlike cross-correlation detects nonlinear and non-monotone dependence.
The Bartlett standard error ±1.96/√N used to declare conventional CCF lags "significant" assumes that at least one series is white noise and that the pair is jointly stationary. Our exposure and PMOA series violate both. They are (i) strongly autocorrelated (bloom persistence over days to weeks), (ii) non-stationary in mean *and* variance (a seasonal trend plus the structural breaks already detected by our changepoint analysis, EnvCpt), and (iii) non-Gaussian, right-skewed and intermittent (spiky). Under these conditions Bartlett's formula for the variance of the sample cross-correlation of two *independent* series reduces to
Var[r_xy(k)] ≈ (1/N) · Σ_j ρ_x(j)·ρ_y(j),
which, when both series are highly autocorrelated, is far larger than the nominal 1/N. Two independent but persistent, co-trending series therefore generate broad, smoothly varying, apparently "significant" CCF lag peaks, the lag-domain form of the Yule (1926) nonsense-correlation and Granger-Newbold (1974) spurious-regression problem, and exactly the "spurious lag structure under non-stationary conditions" flagged by Cheng et al. (2021).
Cross entropy (Sρ) resolves this. First, Sρ is invariant under monotone marginal transformations, so skewness and intermittency do not bias it, and it captures dependence that Pearson misses by construction. Second its significance is taken from a surrogate/sieve-bootstrap null (`surrogate.AR`/`surrogate.ARs`/`surrogate.SA` in `tseriesEntropy`; 500 replicates in our case) that regenerates series preserving each one's own marginal distribution and serial dependence while destroying the cross-dependence. The autocorrelation that inflates the CCF variance is therefore built into the null distribution which is what makes the test valid under precisely the autocorrelated, non-stationary regime that breaks the CCF, and Sρ has further been shown to retain power at small sample sizes (Giannerini et al., 2015; Granger et al., 2004)
Q3b: The wavelet coherence analysis interprets arrows pointing "upward" as phytoplankton leading PMOA. However, the wavelet phase interpretation depends on the convention used in the biwavelet package. Please clarify whether upward arrows correspond to PFT leading PMOA or the reverse, and verify this against known biological timelines.
Response: The convention used here follows the standard biwavelet phase convention and is consistent with the biological timeline we interpret (early-bloom leading signals followed by anti-phase lysis coupling). The convention used is explained in Section 2.6: upward arrows denote phytoplankton leading PMOA and downward arrows PMOA leading phytoplankton, with right = in-phase and left = anti-phase (phytoplankton lysis concurrent with PMOA increase).
Q3c: For the wavelet analysis, you state that "coherence patches exhibiting rapidly rotating or opposing phase arrows within the same significant region were not interpreted." What percentage of significant coherence regions were excluded by this criterion? This should be quantified.
Response: The rotating/opposing-phase criterion follows standard cross-wavelet interpretation, in which phase relationships are assigned physical meaning only where the phase is locally consistent or phase locked within regions of high coherence (Grinsted et al., 2004). This is illustrated with the visibly unstable phase within a single significant patch for PMOA-eBC null case (Figure S4); it is not a counting threshold applied across a population of patches. In the interpreted PFT-PMOA spectra (Figure S21) the significant coherence patches showed coherent, non-rotating phase, so none of the interpreted regions were affected. We clarify in the caption that the criterion excluded only the eBC control among the pairs considered.
- PFT Temporal Dynamics
Q4a: The changepoint analysis shows diatoms entering climax on March 16 and persisting until July 17. However, Figure 2A shows PMOA peaking sharply in early July. Is there a lag between maximum PFT exposure and maximum PMOA? The cross-entropy analysis suggests 2-25 day lags, but the visual inspection of Figure 2A suggests the PMOA peak might lag the PFT peak. Please clarify.
Response: Yes - this lag is the central result of Section 3.4 (now 3.5) and is quantified. Figure 2 shows contemporaneous exposure, whereas the lag between maximum PFT exposure and the PMOA maximum is resolved by the lagged cross-entropy in Figure 4: PMOA responds to diatoms and co-blooming taxa with an early 2-3 day mode and a secondary ~25 day mode, consistent with early labile release followed by lysis/grazing. The sharp early-July PMOA maximum therefore lags the PFT climax onset, exactly as the cross-entropy indicates. We add a forward pointer to the lag analysis where Figure 2 is first discussed in Section 3.3.
Q4b: Table S1 shows diatoms dominating all Longhurst provinces (19.4-65.5%). Given this dominance, why do Prochlorococcus (which are minor contributors) show significant lagged associations with MSA at 30 days? This seems counterintuitive—please discuss whether this is a statistical artifact or a genuine biogeochemical signal.
Response: The apparently disproportionate Prochlorococcus association does not imply a correspondingly large contribution to MSA production. The cross-entropy statistic quantifies normalized pairwise lagged dependence and is therefore not scaled by the absolute biomass of each PFT, allowing a relatively minor taxon to exhibit a detectable temporal association. However, because the PFT fields are strongly covariant, the ~30 d Prochlorococcus feature cannot be uniquely attributed to Prochlorococcus itself and may also reflect its position within broader bloom succession or microbial processing. We therefore interpret this feature as a secondary PFT-associated temporal signature that is consistent with delayed microbial processing of the sulphur pool, including the limited direct DMSP-cleaving capacity reported for Prochlorococcus, but not as evidence of a uniquely resolved Prochlorococcus source of MSA. Its occurrence in MSA, without an equivalent dominant PMOA association, nevertheless supports investigating this pathway specifically within the marine sulphur cycle. This distinction is now added in the lags results and discussion section.
- PMOA-PFT Correlations
Q5a: The correlations during the climax phase (R=0.65-0.70 for diatoms, dinoflagellates, cryptophytes, chlorophytes) are strong. Have you accounted for multiple testing? With 8 PFTs and multiple time lags, the chance of spurious correlations is non-negligible. Please clarify whether p-values were corrected (e.g., FDR or Bonferroni).
Response: This is a fair point. The lagged cross-entropy significances were already derived from 500 surrogate realizations rather than from parametric p-values, but the original analysis did not explicitly control the multiplicity introduced by examining eight phytoplankton functional types (PFTs) over multiple lags. We have therefore revised the inference to control both sources of multiplicity. We did not apply a single-step Bonferroni correction independently to every PFT-by-lag comparison. Bonferroni provides valid family-wise error rate (FWER) control without requiring independence among the tests, but does not exploit their joint dependence structure and can therefore become substantially conservative when test statistics are highly correlated (Ninomiya and Fujisawa, 2007; Proschan and Shaw, 2011). This is particularly relevant to the present lag analysis because neighbouring lag statistics are calculated from extensively overlapping observations and constitute strongly dependent evaluations of the same lag-dependence function. Instead, multiplicity across lags is incorporated directly into the surrogate procedure. For each PFT, the complete cross-entropy function is calculated over the prespecified lag window for every surrogate realization, and the maximum cross-entropy statistic across that window is retained. The observed maximum is then compared with the distribution of surrogate maxima. This tests whether the strongest observed lagged dependence exceeds that expected from searching the complete lag window under the null, while retaining the dependence among neighbouring lag statistics. Such resampling-based multiple-testing approaches explicitly exploit the joint dependence structure of the statistics and provide a substantially better-matched framework than treating strongly dependent comparisons as independent multiplicity contributions (Romano and Wolf, 2005; Westfall and Troendle, 2008).
The resulting eight PFT-level probabilities are subsequently adjusted using the sequential Holm procedure. Holm provides strong FWER control without requiring independence among the hypotheses and is never less powerful than the corresponding single-step Bonferroni procedure (Holm, 1979). We considered false-discovery-rate (FDR) control but did not apply the conventional Benjamini-Hochberg (BH) procedure. The original BH result establishes FDR control under independence (Benjamini and Hochberg, 1995), and its extension to dependent tests requires specific forms of positive regression dependence (Benjamini and Yekutieli, 2001), which cannot be assumed for biologically covarying PFTs. Under more general dependence structures, ordinary Benjamini-Hochberg control is not guaranteed, and anti-conservative behaviour has been demonstrated under strongly correlated testing settings (Kim and Van De Wiel, 2008). This issue is particularly relevant for correlated two-sided tests, for which general Benjamini-Hochberg control under arbitrary correlation has remained theoretically unresolved and has motivated dependence-adjusted procedures (Reiner‐Benaim, 2007; Sarkar and Zhang, 2025). Given the small and prespecified family of eight biological hypotheses, we therefore considered strong FWER control using Holm as the more defensible inferential criterion.
The principal climax-phase associations with diatoms, dinoflagellates, cryptophytes and chlorophytes remain significant under this multiplicity-controlled procedure, whereas weaker associations are interpreted accordingly. Section 2.6 and Figure 4 have been revised to report the corrected inference.
We additionally addressed the covariance among the PFT predictors themselves. Variance inflation factors (VIFs) were calculated for all eight PFTs to quantify the inflation of regression-coefficient variance arising from multicollinearity. The VIF analysis confirms substantial multicollinearity among the PFT predictors. Because fixed VIF thresholds are diagnostic conventions rather than statistically justified rules for selecting a regression estimator, we do not use an arbitrary VIF cutoff to choose between ordinary least squares, ridge regression or LASSO (O’brien, 2007), all eight PFTs are entered simultaneously into an elastic-net regression, with both the mixing parameter α and regularization parameter λ selected by blocked cross-validation. Elastic net continuously spans ridge regression at α = 0 and LASSO at α = 1 and was specifically developed to provide stable regularization and variable selection in the presence of correlated predictors, including a grouping tendency whereby strongly correlated predictors can be retained together (Hoerl and Kennard, 1970; Tibshirani, 1996; Zou and Hastie, 2005). Bootstrap stability is subsequently evaluated to determine whether attribution to individual PFTs is reproducible after their shared covariance has been accounted for. This complementary analysis therefore tests whether individual PFTs retain explanatory information for PMOA once their strong mutual covariance is explicitly incorporated into the multivariable model.
Q5b: Haptophytes show weak correlation (R=0.30) despite being known DMSP producers. Could this be because haptophytes in the North Atlantic are dominated by non-coccolithophorid species that produce less DMSP? Or is this a seasonal effect (haptophyte bloom later in the year)?
Response: We would clarify that R = 0.30 is the haptophyte correlation with PMOA (organic mass), not with MSA; DMSP production is relevant to the sulphur/MSA channel, where haptophytes do in fact appear as early-stage DMS sources (Section 3.4 and Conclusion, point two). For PMOA specifically, the weaker organic link is discussed in the Conclusion: haptophytes emit aliphatics/alcohols mainly under viral lysis rather than during the shelf climax sampled here.
Q5c: The comparison with "August 2009" is interesting but raises questions. August 2009 data used the same instrumentation and PMF methodology? If not, inter-annual differences in PMOA concentrations (one order of magnitude higher) could reflect instrumental changes rather than physical forcing. Please confirm that the same AMS calibration and PMF constraints were applied.
Response: Confirmed. Both the 2009 and 2018 periods were measured with the same HR-ToF-AMS at Mace Head, processed with the same Igor SQUIRREL (v1.65B) / PIKA (v1.25B) chain and the same composition-dependent collection-efficiency and RIE treatment (Section 2.2), and PMOA was constrained with the same Ovadnevaite et al. (2011a) reference spectrum and a-value protocol; MSA used the identical Eq. (1) scaling. The near order-of-magnitude difference between August 2009 and August 2018 is therefore not attributable to instrumental or PMF differences. We add an explicit statement of this methodological continuity where August 2009 is introduced.
Q6a: Cross-entropy shows PMOA lags for diatoms at 2-25 days. How do you distinguish between "early labile release" (2-3 days) and "lysis/grazing" (25 days) given the continuous nature of the signal? Do you have independent evidence (e.g., nutrient data, viral abundance, zooplankton biomass) to support this mechanistic interpretation?
Response: The two modes appear as separate, significant cross-entropy maxima with an early 2-3 day peak and a secondary peak near 25 days (Figure 4A) separated by lower, non-significant intermediate lags.
The revision now brings independent, non-aerosol evidence to bear on the ecological setting of these two modes. Three sets of fields were added along the SQTBA source footprints (Sections 2.4-2.5; Table S5): (i) daily surface macronutrients (NO3, PO4, Si(OH)4) from the Copernicus PISCES-v2 hindcast together with satellite particulate organic carbon (POC); (ii) a literature-constrained, SST-derived virus-host infection potential. Micronekton biomass was already included in the preprint as an independent higher-trophic-level field, providing complementary evidence for trophic processing
We remain explicit about what these constraints are. They are reanalysis, satellite and laboratory-parameterised fields evaluated over the source footprints, not co-located in-situ viral counts, nutrient bottle samples or net-collected zooplankton. Section 3.4 and 3.5 therefore present the labile-versus-lysis assignment as an inference that is now supported by independent biogeochemical and trophic fields.
Q6b: For MSA, you report lags at 4, 9, and 30 days. The 30-day lag for Prochlorococcus seems extraordinarily long. Given typical phytoplankton turnover times of days to weeks, a 30-day lag might reflect seasonal trends rather than causal coupling. Did you test for spurious correlation due to shared seasonality (e.g., detrending the data)?
Response: We agree that the ~30-day Prochlorococcus feature requires caution. The lag analysis was restricted to the spring-summer climax period, and significance was assessed using surrogate realizations preserving each series’ serial dependence. We interpret the 30-day maximum as a secondary PFT-associated signature that may reflect broader bloom succession or delayed microbial processing, as discussed in Q4b.
- Physical Forcing
Q7a: You attribute the August 2009 PMOA enhancement to stronger winds (6.83 m/s median) and cooler SST. However, August 2009 also had higher micronekton biomass (11.2 gC/m² vs. 5.6 gC/m² in 2018). Could the PMOA difference be primarily biological rather than physical? A partial correlation or variance partitioning analysis would help separate biological vs. physical drivers.
Response: We agree, and the physical side of the manuscript has been fully rebuilt for the revision. The air-sea transfer chain is now resolved explicitly over each source footprint: whitecap fraction (Callaghan et al., 2008). the breaking-wave entrained-air volume flux and submicron film-drop fluxes (Deike, 2022), and bubble-plume penetration depth (Cifuentes‐Lorenzen et al., 2023), alongside SST, friction velocity, significant wave height, wave age, wave Reynolds number, atmospheric stability, seawater viscosity and surface tension (Markuszewski et al., 2024; Ovadnevaite et al., 2014). In parallel, the biological side is now resolved by POC, PISCES macronutrients, SST-derived viral potentials and micronekton. All are reported as medians and interquartile ranges for the three footprints in Table S5 and as distributions in Figures S16-S20.
With these drivers, the three case study periods contrasts separate biology from physics more informatively than a partial correlation over three period means could. The first contrast, spring-summer 2018 climax against August 2018, moves physics up and biology down as whitecap fraction, entrained-air flux and significant wave height increase by 54 %, 45 % and 45 % respectively, while the wave Reynolds number increases by only 6 % and the parameterised bubble-plume penetration depth is essentially unchanged (0.77 to 0.72 m). Over the same contrast POC falls by a factor of 1.53 and micronekton exposure by a factor of 1.7, and the strong contemporaneous PFT-PMOA covariance disappears. Stronger mechanical forcing on its own therefore does not sustain PMOA variability, which is the opposite of what a purely physical explanation would predict.
The second contrast, August 2009 against the spring-summer climax is now better constrained. The organic reservoir was essentially the same in the two periods (POC 189 against 191 mg m-3) and the SST-derived viral potentials were lower in August 2009, while every transfer term was far higher: whitecap fraction 4.5-fold, entrained-air flux 5.2-fold, total film-drop number flux 4.0-fold, wave Reynolds number 3.2-fold and penetration depth 2.6-fold. Micronekton exposure was however indeed twice as high. Thus, August 2009 combined comparable PFT concentration and POC with both substantially stronger bubble-mediated transfer and greater higher-trophic-level activity. The observations therefore suggest an important amplification by physical transfer, while the elevated micronekton prevents us from attributing the enhancement to physics alone. As discussed in Q7e, micronekton are interpreted both as active processors of the organic pool through egestion, excretion and mortality and as an indicator of broader higher-trophic-level activity, since micronekton and zooplankton carbon processing covary across Atlantic productivity regimes; these processes can modify the material available for bubble-mediated aerosolisation, with zooplankton grazing and excretion directly linked to enhanced PMOA production (Hernández-León, 2023; Dall’Osto et al., 2025).
Q7b: The wave Reynolds number is used as a proxy for bubble-mediated aerosol production. Did you test alternative parameterizations (e.g., whitecap fraction, bubble plume penetration depth) to see if the conclusions are robust? The Reynolds number depends on assumptions about seawater viscosity and salinity—were these varied in sensitivity tests?
Response: The wave Reynolds number was originally chosen because it integrates wind stress, wave state and seawater viscosity into a single bubble-production proxy previously linked to organic sea-spray enrichment at Mace Head (Ovadnevaite et al., 2014). The original manuscript therefore relied principally on significant wave height and wave Reynolds number to characterise the progression in physical forcing. In the revision, we have substantially expanded this analysis to test whether the interpretation is reproduced by independent bubble-related parameterisations. Specifically, we added modelled whitecap coverage, breaking-wave air-entrainment flux and submicron cap/film-drop production (Deike, 2022), together with a wave-age-constrained estimate of bubble-plume penetration depth (Cifuentes‐Lorenzen et al., 2023). These diagnostics give a consistent physical picture: August 2018 progressed to stronger (relative to the spring-summer climax period) wave breaking, whitecap coverage and air entrainment than the spring-summer climax while parameterised bubble-plume penetration remained essentially unchanged, whereas August 2009 showed the strongest whitecap coverage, air entrainment and film-drop production together with a substantially deeper bubble plume.
On viscosity inputs: salinity in the open North Atlantic source regions varies only within ~34-36 PSU, whose effect on kinematic viscosity is negligible.
- Discussion & Interpretation
Q7a: The statement "diatoms, dinoflagellates, chlorophytes, and cryptophytes all covaried along with PMOA" (line 474-475) is supported by correlation analysis. But correlation does not establish causation. Could the PMOA-PFT correlations simply reflect shared seasonal forcing (e.g., light, temperature, nutrients) rather than direct biological emission? Please discuss how you rule out common environmental drivers.
Response: We agree that contemporaneous correlation alone cannot establish causation, which motivated the additional analyses extending beyond Pearson correlation. Three complementary lines of evidence now support an interpretation involving biologically mediated PMOA variability in addition to shared seasonal forcing. First, the lag structure in Figure 4 resolves the PFT-PMOA relationship into mechanistically distinct early-release and lysis-associated modes, with significance evaluated against surrogate series that preserve the seasonal persistence of the original data (see Q6b). Second, the physical-forcing comparison presented driver by driver in Section 3.4 and Table S5 shows that similar phytoplankton exposure can yield markedly different PMOA responses as air-sea transfer conditions change. In August 2018, stronger physical forcing coincided with reduced PFT-PMOA covariance, demonstrating that the biological association is modulated by the efficiency of ocean-atmosphere transfer and cannot be explained by seasonal co-variation alone. Third, the revision incorporates source-region biogeochemical constraints independent of the aerosol measurements, including macronutrients, POC and SST-derived viral potentials. Temperature and nutrient status are therefore characterised explicitly for each period and incorporated into the interpretation of the observed lag structure. We further state in Section 3.3 that the collinearity and elastic-net analysis support attribution at the level of the biological ensemble, while individual functional-type contributions remain resolved with lower certainty (now toned down in the lags section). Sections 3.4 and the Conclusion now make explicit that the combined temporal, physical and biogeochemical analyses substantially constrain shared-seasonality confounding, while the observational design leaves some residual uncertainty in causal attribution.
Q7b: You invoke "viral lysis" as a mechanism for PMOA release (lines 507-508). Do you have any independent evidence for viral activity during the sampled periods? Without viral abundance data, this remains speculative. Consider tempering this interpretation or citing literature from the same region.
Response: We agree that viral lysis required stronger constraint, and the revision now places this mechanism within an independently characterised ecosystem context.
First, the source-region ecosystem state is quantified using daily surface nitrate, phosphate and silicate from the Copernicus PISCES-v2 hindcast (GLOBAL_MULTIYEAR_BGC_001_029; Aumont et al., 2015) together with satellite particulate organic carbon extracted over the SQTBA footprints (Section 3.4; Table S5). The spring-summer NECS climax contained 1.53-fold more POC than the August 2018 NADR footprint while nitrate and phosphate were strongly drawn down, by factors of 20.6 and 4.9, respectively, and silicate was 25.7% higher. These contrasts provide mechanistic context because nitrogen and phosphorus limitation is known to reduce viral burst size and prolong latent periods in marine phytoplankton (Flynn et al., 2021; Maat and Brussaard, 2016), while silicate limitation is known to facilitate viral infection and accelerate virus-induced mortality in diatoms (Kranzler et al., 2019).
Second, we added a literature-constrained virus-host infection potential derived from SST over the same footprints using experimentally characterised systems representative of the dominant functional types: temperature-dependent Micromonas–prasinovirus production rates (Demory et al., 2017), E. huxleyi-EhV86 burst size (Kendrick et al., 2014), and DNA- and RNA-ctenovirus lysis rates for Chaetoceros tenuissimus (Kimura and Tomaru, 2017). Relative to the NECS climax footprint, the 11.7% warmer August 2018 NADR footprint yielded 32.6% higher Micromonas production potential, 17.2% higher DNA-ctenovirus and 5.4% higher RNA-ctenovirus lysis potentials, while the EhV86 burst-size potential changed by only +3.7%. This spread is consistent with the strong, non-linear and host-specific temperature dependence of marine virus-host interactions (Anesio and Bellas, 2011; Demory et al., 2021; Maat et al., 2017). The combined fields therefore describe a trade-off between host and substrate availability and infection potential. The climax combined abundant hosts and a large organic reservoir with lower modelled viral potential, whereas August 2018 combined higher modelled viral potential with a substantially smaller reservoir, consistent with the host-availability and temperature gradients observed across the North Atlantic (Mojica et al., 2016).
The revised interpretation explicitly treats these SST-derived quantities as literature-constrained potentials at footprint scale, with uncertainty arising from their transfer from laboratory host–virus systems to natural assemblages. Viral lysis is therefore considered alongside grazing as one plausible pathway contributing to delayed post-bloom PMOA formation. The independent nutrient, POC and temperature fields provide the ecological constraints for this interpretation, while the regional viral-ecology literature defines the mechanistic basis and its limits.
Q7c: The discussion of "homeoviscous adaptation" (line 399-401) linking colder SST to increased lipid unsaturation is interesting but speculative for your specific dataset. Did you measure lipid composition in seawater or aerosol? If not, this should be presented as a hypothesis rather than a conclusion.
Response: We agree that the homeoviscous-adaptation mechanism requires careful framing because lipid composition was outside the observational scope of this study. We have therefore revised Section 3.3 and 3.4 to distinguish explicitly between quantities constrained over the source footprints and literature-supported processes used to interpret their combined effects.
Within this broader mechanistic discussion, homeoviscous adaptation is retained as one plausible temperature-sensitive pathway affecting the composition of the organic material available for bubble scavenging. This mechanism is now embedded within a broader discussion of competing SST-sensitive processes in section 3.4. The revised text explicitly distinguishes between processes constrained by the present analysis (e.g. POC, viscosity) and additional temperature-sensitive mechanisms that were not directly measured and are therefore discussed only as plausible, literature-supported pathways (e.g. microbial degradation, bubble residence and dissolution, film drainage and coalescence, surfactant adsorption, molecular transfer to film drops).
Q7d: You note that "not all biological activity is represented (e.g., viruses, archaea, fungi)" (line 518). How significant are these omissions? Recent work suggests archaea and fungi can be important in sea spray aerosol formation. Please quantify or discuss the potential contribution of these missing groups.
Response: We agree that these omissions warrant a more explicit assessment, although their implications differ among viruses, fungi and archaea. Viral activity is now partly constrained in the revision through literature-derived, SST-dependent infection and lysis potentials together with source-region nutrients and POC, as discussed in response to Q7b. For fungi and archaea, taxon-resolved measurements were unavailable nor were reanalyses/satellites or models. We therefore discuss their potential importance here in greater detail using available evidence on marine biomass, phytoplankton mortality and organic-matter processing, while keeping the corresponding manuscript addition concise given the already broad scope and length of the study.
Both archaeae and fungi represent an important part of the carbon pool. The global synthesis of (Bar-On and Milo (2019) estimated approximately 0.3 Gt C each for marine fungi and archaea, and a recent direct reassessment of pelagic fungi obtained 0.32 Gt C, with a 0.19-0.46 Gt C confidence interval (Breyer et al., 2025), These values establish their potential ecosystem relevance, although global standing biomass cannot be converted directly into a fraction of surface POC or PMOA in our source footprints. its taxonomic composition.
Their likely roles also differ mechanistically. For archaea, current evidence points primarily to post-bloom organic-matter transformation and remineralisation. Euryarchaea have transiently accounted for approximately 40% of prokaryotes following a phytoplankton bloom, while Marine Group II archaea can comprise up to approximately 30% of the microbial community after bloom termination and possess genomic potential for degrading proteins, carbohydrates and lipids (Needham and Fuhrman, 2016; Tully, 2019). Marine Group II abundances also covary strongly with several phytoplankton groups and are associated with algae-derived organic matter (Chen et al., 2022). We therefore consider archaeal processing potentially relevant to the composition and persistence of the organic reservoir available for bubble scavenging, while current evidence provides little basis for treating archaea as a major direct phytoplankton-mortality pathway comparable to grazing or viral lysis.
Fungi provide a more direct potential link to phytoplankton mortality and post-bloom carbon processing. Chytrid parasites have been observed infecting marine Skeletonema, Thalassiosira and Chaetoceros populations, with infection tracking host density (Gutiérrez et al., 2016). Experimental work further shows that fungal infection can redirect phytoplankton-derived carbon through the “fungal shunt”, increasing its accessibility to associated heterotrophs and altering the fate of particulate organic matter (Klawonn et al., 2021). Fungal parasitism could therefore contribute to bloom decline and subsequent organic-matter processing in addition to viral lysis and grazing. Its magnitude however remains much less constrained at basin scale: unlike viral lysis and microzooplankton grazing, which have been measured concurrently across North Atlantic gradients (Mojica et al., 2016), comparable regional fungal mortality rates are currently unavailable. We consequently mention the “fungal shot” as an additional unresolved plausible post-bloom pathway.
Q7e: The conclusion states that "this study provides observational evidence that higher trophic levels contribute directly to PMOA" (line 499). Given that micronekton are not primary producers but consumers, is the correlation with micronekton simply tracking the same seasonal signal as phytoplankton? A partial correlation controlling for total PFT biomass would help establish whether micronekton add explanatory power beyond phytoplankton.
Response: This is a valid concern and complements the Physical-Forcing Q7a above. We agree that the observed micronekton association should not be interpreted as demonstrating a uniquely identifiable direct PMOA source. A complete separation of phytoplankton, higher-trophic-level and air-sea transfer contributions to PMOA is beyond the scope of this study. A more complete separation of phytoplankton, zooplankton, micronekton and physical-transfer influences will require longer observations spanning repeated seasonal cycles and ecosystem states. We identify this as future work using the 2008-2019 PMOA record at Mace Head (Moschos et al., 2026) particularly as improved phytoplankton community products and PACE-enabled developments in the long-term ocean-colour record become available.
Accordingly, we have revised the conclusion to avoid the statement that micronekton contribute “directly” to PMOA. This more cautious interpretation remains mechanistically supported by evidence that micronekton transform epipelagic organic carbon through egestion, excretion and mortality (Thibault et al., 2025) that zooplankton and micronekton carbon processing covary across Atlantic productivity regimes (Hernández-León, 2023), with recent observations around the Iberian Peninsula further showing substantial micronekton contributions to active carbon flux in the Atlantic (Couret et al., 2025) and, most directly, that zooplankton grazing and excretion have been associated with substantial enrichment of humic-like organic material in PMOA (Dall’Osto et al., 2025). We therefore interpret the micronekton relationship as evidence that trophic processing contains information relevant to PMOA variability, without assigning an exact fractional source contribution to micronekton themselves.
We still note however that the cross-entropy already places micronekton on partly distinct lag timescales (2, 10 and 30 days; Figure 4C) from the phytoplankton peaks, which is difficult to explain by simple co-seasonality and points to genuine trophic processing (i.e. bubble-mediated enrichment of lipid-rich faecal pellets and excretion products).
- Technical/Editorial Issues
Q8.1: There are duplicate author affiliations and repeated text in the author list (lines 5-8 appear duplicated). Please correct.
Response: We have checked the submitted manuscript: the author list contains a single author line and a single affiliation '1' (School of Natural Sciences, Ryan Institute's C-CAPS, University of Galway), with one present-address footnote for W. Xu; there is no duplicated affiliation or repeated author text. The duplication may have appeared in the reviewer's rendered copy during file handling; we will verify the final typeset proof is clean.
Q8.2: The abbreviation "SPM" (suspended particulate matter) is introduced but not used in the main analysis. Was SPM considered? If not, remove from methods.
Response: SPM is indeed used - it is examined in Figure S22 (panels C and D) alongside chlorophyll-a as an alternative ocean-colour proxy for PMOA, and like raw chl-a it performs poorly comparatively than the PFT exposure during the oligotrophic summer-climax bloom (unless Hampel filtering is applied to remove chl-a artefacts). To remove any ambiguity, we have added a cross-reference to Figure S16 where SPM is first mentioned in the Methods, so it is clear the variable is retrieved and evaluated.
Q8.3: Several references in the text are incomplete or incorrectly formatted (e.g., "Daey and Wakeham_1986" in line 433, "Alberne et al., 2024" in references has no initials). Please check all references against the EGU reference style.
Response: we cannot see these format issues in our version of the submitted file but will nonetheless run a final pass of all references against the Copernicus/EGU style.
Q8.4: The manuscript uses both "micronekton" and "microenokton" (typo on line 165). Please standardize spelling throughout.
Response: 'Micronekton' is now spelled consistently throughout the main text and supplement (22 occurrences plus capitalised sentence-initial forms).
Q8.5: Figure numbers: Figure S5, S6 etc. are referenced but the supplement file was not provided in the review. Please ensure all supplementary figures are properly labeled and referenced.
Response: The Supplementary Information containing Figures S1-S22 and Tables S1-S2, all labelled and cross-referenced from the main text, was submitted with the manuscript. We will ensure that all supplementary figures are properly labelled and referenced.
References
Anesio, A. M. and Bellas, C. M.: Are low temperature habitats hot spots of microbial evolution driven by viruses?, Trends in Microbiology, 19, 52–57, https://doi.org/10.1016/j.tim.2010.11.002, 2011.
Aumont, O., Ethé, C., Tagliabue, A., Bopp, L., and Gehlen, M.: PISCES-v2: an ocean biogeochemical model for carbon and ecosystem studies, Geosci. Model Dev., 8, 2465–2513, https://doi.org/10.5194/gmd-8-2465-2015, 2015.
Bar-On, Y. M. and Milo, R.: The Biomass Composition of the Oceans: A Blueprint of Our Blue Planet, Cell, 179, 1451–1454, https://doi.org/10.1016/j.cell.2019.11.018, 2019.
Benjamini, Y. and Hochberg, Y.: Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing, Journal of the Royal Statistical Society Series B: Statistical Methodology, 57, 289–300, https://doi.org/10.1111/j.2517-6161.1995.tb02031.x, 1995.
Benjamini, Y. and Yekutieli, D.: The control of the false discovery rate in multiple testing under dependency, Ann. Statist., 29, https://doi.org/10.1214/aos/1013699998, 2001.
Breyer, E., Stix, C., Kilker, S., Roller, B. R. K., Panagou, F., Doebke, C., Amano, C., Saavedra, D. E. M., Coll-García, G., Steger-Mähnert, B., Dachs, J., Berrojalbiz, N., Vila-Costa, M., Sobrino, C., Fuentes-Lema, A., Berthiller, F., Polz, M. F., and Baltar, F.: The contribution of pelagic fungi to ocean biomass, Cell, 188, 3992-4002.e13, https://doi.org/10.1016/j.cell.2025.05.004, 2025.
Callaghan, A., De Leeuw, G., Cohen, L., and O’Dowd, C. D.: Relationship of oceanic whitecap coverage to wind speed and wind history, Geophysical Research Letters, 35, 2008GL036165, https://doi.org/10.1029/2008GL036165, 2008.
Chen, S., Tao, J., Chen, Y., Wang, W., Fan, L., and Zhang, C.: Interactions Between Marine Group II Archaea and Phytoplankton Revealed by Population Correlations in the Northern Coast of South China Sea, Front. Microbiol., 12, 785532, https://doi.org/10.3389/fmicb.2021.785532, 2022.
Cheng, Y., Hui, Y., McAleer, M., and Wong, W.-K.: Spurious Relationships for Nearly Non-Stationary Series, JRFM, 14, 366, https://doi.org/10.3390/jrfm14080366, 2021.
Chevassus, E., Fossum, K. N., Ceburnis, D., Lei, L., Lin, C., Xu, W., O’Dowd, C., and Ovadnevaite, J.: Marine organic aerosol at Mace Head: effects from phytoplankton and source region variability, Atmos. Chem. Phys., 25, 4107–4129, https://doi.org/10.5194/acp-25-4107-2025, 2025.
Cifuentes‐Lorenzen, A., Zappa, C. J., Randolph, K., and Edson, J. B.: Scaling the Bubble Penetration Depth in the Ocean, JGR Oceans, 128, e2022JC019582, https://doi.org/10.1029/2022JC019582, 2023.
Couret, M., Sarmiento-Lezcano, A. N., Landeira, J. M., Giering, S. L. C., Major, W., Olivar, M. P., Díaz-Pérez, J., Castellón, A., and Hernández-León, S.: Zooplankton and micronekton active flux around the Iberian Peninsula, Front. Mar. Sci., 12, 1652483, https://doi.org/10.3389/fmars.2025.1652483, 2025.
Dall’Osto, M., Schmidt, K., Campbell, R. G., Nomura, D., Park, J., Yoon, Y. J., and Park, J.: Zooplankton grazing increases atmospheric primary aerosol production in the high Arctic, Elem Sci Anth, 13, 00078, https://doi.org/10.1525/elementa.2024.00078, 2025.
Deike, L.: Mass Transfer at the Ocean–Atmosphere Interface: The Role of Wave Breaking, Droplets, and Bubbles, Annu. Rev. Fluid Mech., 54, 191–224, https://doi.org/10.1146/annurev-fluid-030121-014132, 2022.
Demory, D., Arsenieff, L., Simon, N., Six, C., Rigaut-Jalabert, F., Marie, D., Ge, P., Bigeard, E., Jacquet, S., Sciandra, A., Bernard, O., Rabouille, S., and Baudoux, A.-C.: Temperature is a key factor in Micromonas –virus interactions, The ISME Journal, 11, 601–612, https://doi.org/10.1038/ismej.2016.160, 2017.
Demory, D., Weitz, J. S., Baudoux, A., Touzeau, S., Simon, N., Rabouille, S., Sciandra, A., and Bernard, O.: A thermal trade‐off between viral production and degradation drives virus‐phytoplankton population dynamics, Ecology Letters, 24, 1133–1144, https://doi.org/10.1111/ele.13722, 2021.
Flynn, K. J., Kimmance, S. A., Clark, D. R., Mitra, A., Polimene, L., and Wilson, W. H.: Modelling the Effects of Traits and Abiotic Factors on Viral Lysis in Phytoplankton, Front. Mar. Sci., 8, 667184, https://doi.org/10.3389/fmars.2021.667184, 2021.
Giannerini, S. and Goracci, G.: Entropy-Based Tests for Complex Dependence in Economic and Financial Time Series with the R Package tseriesEntropy, Mathematics, 11, 757, https://doi.org/10.3390/math11030757, 2023.
Giannerini, S., Maasoumi, E., and Dagum, E. B.: Entropy testing for nonlinear serial dependence in time series, Biometrika, 102, 661–675, https://doi.org/10.1093/biomet/asv007, 2015.
Granger, C. W., Maasoumi, E., and Racine, J.: A Dependence Metric for Possibly Nonlinear Processes, Journal Time Series Analysis, 25, 649–669, https://doi.org/10.1111/j.1467-9892.2004.01866.x, 2004.
Granger, C. W. J. and Newbold, P.: Spurious regressions in econometrics, Journal of Econometrics, 2, 111–120, https://doi.org/10.1016/0304-4076(74)90034-7, 1974.
Grinsted, A., Moore, J. C., and Jevrejeva, S.: Application of the cross wavelet transform and wavelet coherence to geophysical time series, Nonlin. Processes Geophys., 11, 561–566, https://doi.org/10.5194/npg-11-561-2004, 2004.
Gutiérrez, M. H., Jara, A. M., and Pantoja, S.: Fungal parasites infect marine diatoms in the upwelling ecosystem of the Humboldt current system off central Chile, Environmental Microbiology, 18, 1646–1653, https://doi.org/10.1111/1462-2920.13257, 2016.
Hernández-León, S.: The biological carbon pump, diel vertical migration, and carbon dioxide removal, iScience, 26, 107835, https://doi.org/10.1016/j.isci.2023.107835, 2023.
Hoerl, A. E. and Kennard, R. W.: Ridge Regression: Biased Estimation for Nonorthogonal Problems, Technometrics, 12, 55–67, https://doi.org/10.1080/00401706.1970.10488634, 1970.
Junger, W. L. and Ponce De Leon, A.: Imputation of missing data in time series for air pollutants, Atmospheric Environment, 102, 96–104, https://doi.org/10.1016/j.atmosenv.2014.11.049, 2015.
Kendrick, B. J., DiTullio, G. R., Cyronak, T. J., Fulton, J. M., Van Mooy, B. A. S., and Bidle, K. D.: Temperature-Induced Viral Resistance in Emiliania huxleyi (Prymnesiophyceae), PLoS ONE, 9, e112134, https://doi.org/10.1371/journal.pone.0112134, 2014.
Kim, K. I. and Van De Wiel, M. A.: Effects of dependence in high-dimensional multiple testing problems, BMC Bioinformatics, 9, 114, https://doi.org/10.1186/1471-2105-9-114, 2008.
Kimura, K. and Tomaru, Y.: Effects of temperature and salinity on diatom cell lysis by DNA and RNA viruses, Aquat. Microb. Ecol., 79, 79–83, https://doi.org/10.3354/ame01818, 2017.
Klawonn, I., Van Den Wyngaert, S., Parada, A. E., Arandia-Gorostidi, N., Whitehouse, M. J., Grossart, H.-P., and Dekas, A. E.: Characterizing the “fungal shunt”: Parasitic fungi on diatoms affect carbon flow and bacterial communities in aquatic microbial food webs, Proc. Natl. Acad. Sci. U.S.A., 118, e2102225118, https://doi.org/10.1073/pnas.2102225118, 2021.
Kranzler, C. F., Krause, J. W., Brzezinski, M. A., Edwards, B. R., Biggs, W. P., Maniscalco, M., McCrow, J. P., Van Mooy, B. A. S., Bidle, K. D., Allen, A. E., and Thamatrakoln, K.: Silicon limitation facilitates virus infection and mortality of marine diatoms, Nat Microbiol, 4, 1790–1797, https://doi.org/10.1038/s41564-019-0502-x, 2019.
Maat, D. and Brussaard, C.: Both phosphorus- and nitrogen limitation constrain viral proliferation in marine phytoplankton, Aquat. Microb. Ecol., 77, 87–97, https://doi.org/10.3354/ame01791, 2016.
Maat, D., Biggs, T., Evans, C., Van Bleijswijk, J., Van Der Wel, N., Dutilh, B., and Brussaard, C.: Characterization and Temperature Dependence of Arctic Micromonas polaris Viruses, Viruses, 9, 134, https://doi.org/10.3390/v9060134, 2017.
Markuszewski, P., Nilsson, E. D., Zinke, J., Mårtensson, E. M., Salter, M., Makuch, P., Kitowska, M., Niedźwiecka-Wróbel, I., Drozdowska, V., Lis, D., Petelski, T., Ferrero, L., and Piskozub, J.: Multi-year gradient measurements of sea spray fluxes over the Baltic Sea and the North Atlantic Ocean, https://doi.org/10.5194/egusphere-2024-1254, 3 June 2024.
Mojica, K. D. A., Huisman, J., Wilhelm, S. W., and Brussaard, C. P. D.: Latitudinal variation in virus-induced mortality of phytoplankton across the North Atlantic Ocean, The ISME Journal, 10, 500–513, https://doi.org/10.1038/ismej.2015.130, 2016.
Moritz, S. and Bartz-Beielstein, T.: imputeTS: Time Series Missing Value Imputation in R, The R Journal, 9, 207–218, https://doi.org/10.32614/RJ-2017-009, 2017.
Moschos, V., Chevassus, E., Fossum, K., Lei, L., Pernov, J., Humphries, R., Keywood, M., O’Dowd, C., Ceburnis, D., and Ovadnevaite, J.: Rising natural aerosols drive marine radiative forcing as pollution declines, https://doi.org/10.21203/rs.3.rs-8776460/v1, 4 March 2026.
Needham, D. M. and Fuhrman, J. A.: Pronounced daily succession of phytoplankton, archaea and bacteria following a spring bloom, Nat Microbiol, 1, 16005, https://doi.org/10.1038/nmicrobiol.2016.5, 2016.
Ninomiya, Y. and Fujisawa, H.: A Conservative Test for Multiple Comparison Based on Highly Correlated Test Statistics, Biometrics, 63, 1135–1142, https://doi.org/10.1111/j.1541-0420.2007.00821.x, 2007.
O’brien, R. M.: A Caution Regarding Rules of Thumb for Variance Inflation Factors, Qual Quant, 41, 673–690, https://doi.org/10.1007/s11135-006-9018-6, 2007.
Ovadnevaite, J., Manders, A., de Leeuw, G., Ceburnis, D., Monahan, C., Partanen, A.-I., Korhonen, H., and O’Dowd, C. D.: A sea spray aerosol flux parameterization encapsulating wave state, Atmos. Chem. Phys., 14, 1837–1852, https://doi.org/10.5194/acp-14-1837-2014, 2014.
Proschan, M. A. and Shaw, P. A.: Asymptotics of Bonferroni for dependent normal test statistics, Statistics & Probability Letters, 81, 739–748, https://doi.org/10.1016/j.spl.2010.11.013, 2011.
Reiner‐Benaim, A.: FDR Control by the BH Procedure for Two‐Sided Correlated Tests with Implications to Gene Expression Data Analysis, Biometrical J, 49, 107–126, https://doi.org/10.1002/bimj.200510313, 2007.
Romano, J. P. and Wolf, M.: Exact and Approximate Stepdown Methods for Multiple Hypothesis Testing, Journal of the American Statistical Association, 100, 94–108, https://doi.org/10.1198/016214504000000539, 2005.
Sarkar, S. K. and Zhang, S.: Shifted BH methods for controlling false discovery rate in multiple testing of the means of correlated normals against two-sided alternatives, Journal of Statistical Planning and Inference, 236, 106238, https://doi.org/10.1016/j.jspi.2024.106238, 2025.
Thibault, H., Ménard, F., Abitbol-Spangaro, J., Poggiale, J.-C., and Martini, S.: Modeling the contribution of micronekton diel vertical migrations to carbon export in the mesopelagic zone, Biogeosciences, 22, 2181–2200, https://doi.org/10.5194/bg-22-2181-2025, 2025.
Tibshirani, R.: Regression Shrinkage and Selection Via the Lasso, Journal of the Royal Statistical Society Series B: Statistical Methodology, 58, 267–288, https://doi.org/10.1111/j.2517-6161.1996.tb02080.x, 1996.
Tully, B. J.: Metabolic diversity within the globally abundant Marine Group II Euryarchaea offers insight into ecological patterns, Nat Commun, 10, 271, https://doi.org/10.1038/s41467-018-07840-4, 2019.
Westfall, P. H. and Troendle, J. F.: Multiple Testing with Minimal Assumptions, Biometrical J, 50, 745–755, https://doi.org/10.1002/bimj.200710456, 2008.
Wijesekara, W. M. L. K. N. and Liyanage, L.: Comparison of Imputation Methods for Missing Values in Air Pollution Data: Case Study on Sydney Air Quality Index, in: Advances in Information and Communication, vol. 1130, edited by: Arai, K., Kapoor, S., and Bhatia, R., Springer International Publishing, Cham, 257–269, https://doi.org/10.1007/978-3-030-39442-4_20, 2020.
Yan, S., Xu, G., Zhang, H., Wang, J., Xu, F., Gao, X., Zhang, J., Wu, J., and Yang, G.: Factors Controlling DMS Emission and Atmospheric Sulfate Aerosols in the Western Pacific Continental Sea, JGR Oceans, 129, e2024JC020886, https://doi.org/10.1029/2024JC020886, 2024.
Yule, G. U.: Why do we Sometimes get Nonsense-Correlations between Time-Series?--A Study in Sampling and the Nature of Time-Series, Journal of the Royal Statistical Society, 89, 1, https://doi.org/10.2307/2341482, 1926.
Zou, H. and Hastie, T.: Regularization and Variable Selection Via the Elastic Net, Journal of the Royal Statistical Society Series B: Statistical Methodology, 67, 301–320, https://doi.org/10.1111/j.1467-9868.2005.00503.x, 2005.
Citation: https://doi.org/10.5194/egusphere-2026-2745-AC2
-
AC2: 'Reply on RC1', Emmanuel Chevassus, 20 Aug 2026
-
RC2: 'Comment on egusphere-2026-2745', Anonymous Referee #2, 01 Jul 2026
The manuscript deals with primary marine organic aerosol formation and its association with plankton abundance and biodiversity in the eastern North Atlantic. By leveraging new comprehensive oceanographic datasets on phytoplankton types and micronecton, and by integrating the analysis of air-sea interfacial physical processes with a detailed investigation of the potentially relevant ecological processes in the surface ocean, this study clearly provides advances in current research on aerosol-biosphere interactions. The scientific approach is rather straightforward, while the discussion of the results is sometimes hard to follow, especially for an “atmospheric” readership. The paper provides convincing evidence of the complexity of the potential biological processes involved. At the same time, the reader may get the impression that this complexity is such that it is not actually constrained by the data. For instance, (line 384) an increase of viscosity is associated with a slower degradation of surfactants and with a higher degree of lipid supersaturation, both stimulating PMOA production; however, a higher viscosity would decrease the buoyancy of small bubbles and their resurfacing. I understand the Authors’ efforts to support the statistical analysis with plausible mechanisms, however, when contrasting effects are expected, this should be acknowledged. Alternatively, the Authors should simplify the discussion when too speculative or based on ad hoc hypotheses.
I have one major comments on the results. In Section 3.3, it is highlighted that “During the spring-summer climax bloom, PMOA closely tracks recent biological activity”, as clearly witnessed by the data shown in Fig. 2A, while in Section 3.4, the Authors observed that “For primary marine organic aerosol (PMOA) during the spring-summer climax phase, PMOA-PFT cross entropy shows significant lags between PMOA and several PFT” with lag values up to 25 days or more (Fig. 4). Reasons for this lagged association were found in procaryote-diatoms interactions or other trophic-level processes. These findings discussed in Section 3.4 contrast with the observations done in Section 3.3. I wonder whether the lagged association peaking at > 20 days might reflect the succession of two blooms (and two PMOA peaks) during the spring-summer maximum (Fig. 2A): the second PMOA peak would positively correlate with the first algal bloom with a lagged association of > 20 days. This is however only apparent, because if not, the second algal bloom would lead to a third PMOA peak at the end of July which did not take place in fact (Fig. S13).
The Conclusions section does not provide a fair account of the main findings of this study because too many statements about the ecological processed driving the variability in PMOA are just unsupported and too speculative. For instance, the simple finding that “During the spring-summer climax, diatoms, dinoflagellates, chlorophytes, and cryptophytes all covaried along with PMOA” is discussed in terms of the diverse PMOA precursors that can be produced by the specific PFTs leading the Authors to conclude that a cumulative contribution from all major PFTs to PMOA in this period of the year is “expected” (line 475). Surprisingly, the Authors neglect the most obvious explanation: all PFTs covary with PMOA because they first covary between each other. If so (can a table be provided?), it remains unclear whether PMOA are driven mainly from some individual PFTs rather than from their ensemble.
The paragraph discussing the lagged response of PMOA during August 2018 and 2009 in contrast with the closer coupling found in spring- early summer (lines 500 – 507) is clear and useful, but again I wonder if the indicative mood is appropriate in sentences like “cooler SST raised viscosity and stabilised gel-like surfactants, higher significant wave heights and a markedly larger wave Reynolds number sustained breaking waves and film drops so the same PFT pool was transferred to the atmosphere far more efficiently”. Are any data about the physical state of gel-like surfactants and on film drops fluxes available in this study?
Specific comments:
Figure 2A. The colour palette does not allow to clearly associate the lines in the plot with the individual PFTs.
Figure S12: what do “A”, “B”, “C” stand for?
Citation: https://doi.org/10.5194/egusphere-2026-2745-RC2 -
AC1: 'Reply on RC2', Emmanuel Chevassus, 20 Aug 2026
We thank Reviewer 2 for the careful reading and constructive comments. Each point is addressed below (author responses in italic).
The manuscript deals with primary marine organic aerosol formation and its association with plankton abundance and biodiversity in the eastern North Atlantic. By leveraging new comprehensive oceanographic datasets on phytoplankton types and micronekton, and by integrating the analysis of air-sea interfacial physical processes with a detailed investigation of the potentially relevant ecological processes in the surface ocean, this study clearly provides advances in current research on aerosol-biosphere interactions. The scientific approach is rather straightforward, while the discussion of the results is sometimes hard to follow, especially for an “atmospheric” readership. The paper provides convincing evidence of the complexity of the potential biological processes involved. At the same time, the reader may get the impression that this complexity is such that it is not actually constrained by the data. For instance, (line 384) an increase of viscosity is associated with a slower degradation of surfactants and with a higher degree of lipid supersaturation, both stimulating PMOA production; however, a higher viscosity would decrease the buoyancy of small bubbles and their resurfacing. I understand the Authors’ efforts to support the statistical analysis with plausible mechanisms, however, when contrasting effects are expected, this should be acknowledged. Alternatively, the Authors should simplify the discussion when too speculative or based on ad hoc hypotheses.
Response: We agree with the reviewer that the previous discussion could give an overly deterministic impression where several biological and physical processes may exert contrasting effects. We have therefore substantially rewritten the corresponding discussion, now Section 3.4 (“Source regions, air-sea physics and nutrients”), to distinguish more clearly between processes constrained by our analysis and mechanistic interpretations. In particular, the revised text now explicitly acknowledges the competing effects highlighted by the reviewer: increased viscosity may preserve surface-active material and retard bubble-film drainage and coalescence, while simultaneously reducing the rise velocity of small bubbles and increasing their dissolution before reaching the surface, potentially offsetting this enhanced preservation and scavenging. We therefore no longer interpret SST or viscosity as having a unidirectional effect on primary marine organic aerosol (PMOA) production.More broadly, the revised section now evaluates the three periods through the combined constraints of surface particulate organic carbon (POC), nutrients and SST-derived viral parameters together with wave Reynolds number, whitecap coverage, air entrainment, bubble-plume penetration and parameterised film-drop production. This comparison shows that stronger mechanical forcing in August 2018 alone was insufficient to maintain the strong PFT-PMOA coupling observed during the spring–summer climax, whereas August 2009 combined a substantial organic reservoir with much stronger bubble-mediated transfer. The discussion has consequently been reframed around the balance between biological source strength and physical transfer, with mechanisms that remain inferential explicitly presented as possible explanations.
We have also systematically moderated causal language throughout the manuscript. To further constrain how much of the observed PMOA variability can be explained from surface ocean biology, we additionally introduced a joint elastic-net regression in which all eight PFTs and micronekton are entered simultaneously. Regularisation parameters were selected using blocked cross-validation and attribution stability was assessed with moving-block bootstrap resampling, thereby accounting explicitly for the strong covariance among the biological predictors while also demonstrating that a substantial fraction of PMOA variability remains unresolved by contemporaneous biological exposure alone. A fully resolved source function for PMOA variability jointly resolving surface ocean biology, lags and air–sea transfer physics is beyond the scope of the present study and will require longer observations spanning repeated seasonal cycles and ecosystem states. We identify this as future work using the 2008-2019 PMOA/MSA record at Mace Head (Moschos et al., 2026) and through mesocosms/modelling work.
We also acknowledge the reviewer’s comment regarding accessibility for an atmospheric readership. We have attempted to remove some redundant ecological terms (e.g. ; commensals), but the underlying ecological complexity was not reduced, since an important result of this study is precisely that phytoplankton functional types, lysis, grazing and nutrients processing jointly shape the production transforming and timing of marine organic aerosols. We have therefore focused the revision on making the distinction between observational constraints and mechanistic interpretation more explicit throughout the results and discussion.
I have one major comments on the results. In Section 3.3, it is highlighted that “During the spring-summer climax bloom, PMOA closely tracks recent biological activity”, as clearly witnessed by the data shown in Fig. 2A, while in Section 3.4, the Authors observed that “For primary marine organic aerosol (PMOA) during the spring-summer climax phase, PMOA-PFT cross entropy shows significant lags between PMOA and several PFT” with lag values up to 25 days or more (Fig. 4). Reasons for this lagged association were found in procaryote-diatoms interactions or other trophic-level processes. These findings discussed in Section 3.4 contrast with the observations done in Section 3.3. I wonder whether the lagged association peaking at > 20 days might reflect the succession of two blooms (and two PMOA peaks) during the spring-summer maximum (Fig. 2A): the second PMOA peak would positively correlate with the first algal bloom with a lagged association of > 20 days. This is however only apparent, because if not, the second algal bloom would lead to a third PMOA peak at the end of July which did not take place in fact (Fig. S13).
Response: We agree that this is an important alternative interpretation of the long-lag feature and that the original text did not sufficiently distinguish the robust short-lag response from the more ambiguous secondary association. We would first clarify that Sections 3.3 and 3.4 (now 3.5) describe different aspects of the same relationship. Section 3.3 quantifies the strong contemporaneous covariance between PMOA and the dominant PFTs over the spring-summer climax, whereas the cross-entropy and biwavelets analyses resolve the temporal structure within that covariance. For the dominant PFTs, the latter shows a pronounced short-lag mode at approximately 2-3 d together with a secondary feature near 25 d, separated by substantially weaker intermediate dependence. Thus, the close coupling described in Section 3.3 is principally compatible with the short response identified in Section 3.6. We have clarified this connection in the revised text.
We nevertheless agree with the reviewer that the secondary >20 d feature cannot be uniquely assigned to lysis, grazing or prokaryote-diatom interactions. The spring-summer record contains successive PFT-PMOA episodes separated by a similar timescale, so shifting the series by approximately 20-30 d can partly align the later PMOA episode with earlier bloom variability. At the broader seasonal scale, however, the changepoint analysis (Figure S5) supports an extended spring-summer climax and does not identify a structural break separating two distinct diatom blooms: diatoms remain within the same climax regime from 16 March until 17 July, while the other major PFTs enter the climax sequentially during spring. The apparent two-peaks structure in Figure 2A can therefore also represent internal variability within one prolonged bloom.
Our surrogate procedure preserves the serial dependence of each time series, and the revised lag-window inference additionally controls the multiplicity associated with searching across lags and PFTs. This establishes that the observed lagged dependence exceeds that expected from independently persistent time series, but it cannot determine whether a significant long-lag maximum arises from delayed biological processing or from the temporal spacing of successive episodes within the same bloom. We also agree that the absence of a corresponding third PMOA maximum following the later PFT increase argues against interpreting ~25 d as a fixed or recurrent PMOA response time.
We have therefore revised the lags estimate section to separate statistical detection from mechanistic attribution. The 2-3 d mode is retained as the principal evidence for rapid PFT-PMOA coupling, while the ~25 d feature is now described as a secondary long-lag association that may reflect post-bloom microbial or trophic processing, viral/fungal shunt and/or the temporal succession of PFT-PMOA episodes within the prolonged bloom. The nutrient, POC, SST-derived viral-potential and micronekton fields introduced in the reply to reviewer 1 further provide independent ecological support for the plausibility of a delayed biological processing from viral lysis, but they do not uniquely attribute the ~25 d feature to that mechanism. We have consequently removed wording describing this feature as a distinct “lysis-stage” response and moderated the corresponding mechanistic interpretation throughout the manuscript.
The Conclusions section does not provide a fair account of the main findings of this study because too many statements about the ecological processed driving the variability in PMOA are just unsupported and too speculative. For instance, the simple finding that “During the spring-summer climax, diatoms, dinoflagellates, chlorophytes, and cryptophytes all covaried along with PMOA” is discussed in terms of the diverse PMOA precursors that can be produced by the specific PFTs leading the Authors to conclude that a cumulative contribution from all major PFTs to PMOA in this period of the year is “expected” (line 475). Surprisingly, the Authors neglect the most obvious explanation: all PFTs covary with PMOA because they first covary between each other. If so (can a table be provided?), it remains unclear whether PMOA are driven mainly from some individual PFTs rather than from their ensemble.
Response: We agree with the reviewer that covariance among the PFTs is the first issue that must be resolved before interpreting their individual relationships with PMOA. This point was also raised by Reviewer 1 and has now been addressed explicitly in the revised manuscript. We first quantified multicollinearity among all eight PFT exposure fields and micronekton using variance inflation factors (VIFs). The analysis confirms strong covariance among the PFTs: VIFs range from 16.8 to 161.6 for the eight PFTs, with pelagophytes, chlorophytes, haptophytes and cryptophytes showing particularly strong collinearity, while micronekton is much less affected (VIF = 3.3). Thus, the reviewer is correct that the strong correlations of several PFTs with PMOA cannot be interpreted as independent evidence. To determine whether PMOA variability could nevertheless be attributed preferentially to individual PFTs after accounting for this shared covariance, all eight PFTs and micronekton were entered simultaneously into an elastic-net regression, with both α and λ selected by blocked cross-validation and attribution stability evaluated using moving-block bootstrap resampling. This analysis is presented in the revised Section 3.3, Figure S17 and Tables S3-S4. The cross-validation surface was nearly flat across α and the nominal optimum occurred at α = 0, corresponding to a ridge solution retaining all nine biological predictors. Thus, the data provide no statistical preference for a sparse model in which one or a small number of PFTs dominate PMOA variability.
The paragraph discussing the lagged response of PMOA during August 2018 and 2009 in contrast with the closer coupling found in spring- early summer (lines 500 – 507) is clear and useful, but again I wonder if the indicative mood is appropriate in sentences like “cooler SST raised viscosity and stabilised gel-like surfactants, higher significant wave heights and a markedly larger wave Reynolds number sustained breaking waves and film drops so the same PFT pool was transferred to the atmosphere far more efficiently”. Are any data about the physical state of gel-like surfactants and on film drops fluxes available in this study?
Response: We agree that the original wording did not sufficiently distinguish quantities constrained in this study from literature-supported mechanisms used to interpret them. We did not measure the physical state, composition or rheology of gel-like particles or surfactant films in the source waters, and we have therefore revised this mechanism to be explicitly conditional as also addressed in our response to Reviewer 1. As addressed there, temperature-dependent changes in organic-matter persistence, film drainage, coalescence and surfactant behaviour are treated as plausible mechanisms whose net effect cannot be isolated from the present observations.
The physical transfer pathway is, however, now considerably better constrained than in the original manuscript. In the revision, we added source-footprint parameterisations of whitecap coverage, breaking-wave air-entrainment flux, submicron film-drop production following Deike (2022), and bubble-plume penetration depth following Cifuentes-Lorenzen et al. (2023), alongside significant wave height, wave Reynolds number and seawater viscosity. These quantities are reported in Table S5 and Figures S19-S20.
Specific comments:
Figure 2A. The colour palette does not allow to clearly associate the lines in the plot with the individual PFTs.
Response: The figure was edited for visual clarity.
Figure S12: what do “A”, “B”, “C” stand for?
Response: Thanks for pointing out this omission. The figure caption now reads “Figure S12. Distributions of air-mass exposure to micronekton biomass (g C m⁻² of seawater) during the three case-study periods: (A) spring-summer 2018, (B) August 2018, and (C) August 2009. Horizontal bars indicate the medians, and diamonds indicate the geometric means. Values above each violin denote the geometric mean ± geometric standard deviation.”
Citation: https://doi.org/10.5194/egusphere-2026-2745-AC1
-
AC1: 'Reply on RC2', Emmanuel Chevassus, 20 Aug 2026
Interactive computing environment
Marine Organic Aerosol Reflects Ecosystem Variability from Phytoplankton Functional Types to Micronekton Emmanuel Chevassus https://zenodo.org/records/20155093
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 159 | 72 | 26 | 257 | 33 | 23 | 20 |
- HTML: 159
- PDF: 72
- XML: 26
- Total: 257
- Supplement: 33
- BibTeX: 23
- EndNote: 20
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General Assessment
This manuscript presents a comprehensive year-long study combining high-resolution aerosol mass spectrometry with gap-free phytoplankton functional type (PFT) data and HYSPLIT trajectory analysis at the Mace Head station. The work demonstrates that PFT-specific linkages to PMOA and MSA are more nuanced than bulk chlorophyll-a approaches, with distinct lag structures and seasonal shifts in the dominant PFT-aerosol couplings. The methodology is sophisticated, combining PMF source apportionment, cross-entropy lag analysis, and wavelet coherence.
The study makes a valuable contribution to understanding marine aerosol sources, but several methodological and interpretive aspects require clarification before publication.
The following issues should be clarified to improve the manuscript further:
1. PMF Source Apportionment
Q1a: The PMOA factor was constrained using reference spectra from Ovadnevaite et al. (2011a) with a-values ranging from 0.1 to 0.5. How sensitive are your PMOA concentrations to the choice of a-value? You mention that 3-5 factor solutions showed unrealistic behavior, but what about varying a-values within the 6-factor solution? A sensitivity analysis should be presented or referenced in the supplement.
Q1b: The PMOA mass spectrum shows 55% oxygenated carbons and 39% aliphatic fragments. How does this fingerprint compare to the reference spectrum you used for constraint? Is there evidence of significant deviation that might indicate mixing with other sources?
Q1c: You excluded pollution events using MAAP/eBC data. What was the threshold for flagging pollution events, and what percentage of data were excluded? This is critical for ensuring the "clean marine" signature you claim.
2. PFT Data and Exposure
Q2a: The AIGD-PFT dataset has 4 km resolution, but your HYSPLIT trajectories use a 20 km radius for exposure calculations. How do you account for the spatial scale mismatch between PFT patchiness (kilometer-scale features noted on line 252) and trajectory averaging?
Q2b: You filtered trajectories with endpoints below 850 hPa and boundary layer heights below 50 m. What fraction of total trajectories were excluded by these filters?
Q2c: The exposure calculation assumes aerosols are well-mixed within the boundary layer. Did you test the sensitivity to using a different vertical mixing scheme or decay rate for organic aerosols during transport? This is particularly relevant for the 72-hour back trajectories, where loss processes could be significant.
3. Cross-Entropy and Wavelet Analysis
Q3a: Cross-entropy was used instead of conventional cross-correlation to avoid "spurious lag structure under non-stationary conditions." Did you test both methods and compare results? Figure S4 shows the PMOA-eBC comparison failing cross-entropy, but this doesn't demonstrate that cross-entropy is superior—only that it's more conservative. A direct comparison table would strengthen your methodological choice.
Q3b: The wavelet coherence analysis interprets arrows pointing "upward" as phytoplankton leading PMOA. However, the wavelet phase interpretation depends on the convention used in the biwavelet package. Please clarify whether upward arrows correspond to PFT leading PMOA or the reverse, and verify this against known biological timelines.
Q3c: For the wavelet analysis, you state that "coherence patches exhibiting rapidly rotating or opposing phase arrows within the same significant region were not interpreted." What percentage of significant coherence regions were excluded by this criterion? This should be quantified.
4. PFT Temporal Dynamics
Q4a: The changepoint analysis shows diatoms entering climax on March 16 and persisting until July 17. However, Figure 2A shows PMOA peaking sharply in early July. Is there a lag between maximum PFT exposure and maximum PMOA? The cross-entropy analysis suggests 2-25 day lags, but the visual inspection of Figure 2A suggests the PMOA peak might lag the PFT peak. Please clarify.
Q4b: Table S1 shows diatoms dominating all Longhurst provinces (19.4-65.5%). Given this dominance, why do Prochlorococcus (which are minor contributors) show significant lagged associations with MSA at 30 days? This seems counterintuitive—please discuss whether this is a statistical artifact or a genuine biogeochemical signal.
5. PMOA-PFT Correlations
Q5a: The correlations during the climax phase (R=0.65-0.70 for diatoms, dinoflagellates, cryptophytes, chlorophytes) are strong. Have you accounted for multiple testing? With 8 PFTs and multiple time lags, the chance of spurious correlations is non-negligible. Please clarify whether p-values were corrected (e.g., FDR or Bonferroni).
Q5b: Haptophytes show weak correlation (R=0.30) despite being known DMSP producers. Could this be because haptophytes in the North Atlantic are dominated by non-coccolithophorid species that produce less DMSP? Or is this a seasonal effect (haptophyte bloom later in the year)?
Q5c: The comparison with "August 2009" is interesting but raises questions. August 2009 data used the same instrumentation and PMF methodology? If not, inter-annual differences in PMOA concentrations (one order of magnitude higher) could reflect instrumental changes rather than physical forcing. Please confirm that the same AMS calibration and PMF constraints were applied.
6. Lag Estimates
Q6a: Cross-entropy shows PMOA lags for diatoms at 2-25 days. How do you distinguish between "early labile release" (2-3 days) and "lysis/grazing" (25 days) given the continuous nature of the signal? Do you have independent evidence (e.g., nutrient data, viral abundance, zooplankton biomass) to support this mechanistic interpretation?
Q6b: For MSA, you report lags at 4, 9, and 30 days. The 30-day lag for Prochlorococcus seems extraordinarily long. Given typical phytoplankton turnover times of days to weeks, a 30-day lag might reflect seasonal trends rather than causal coupling. Did you test for spurious correlation due to shared seasonality (e.g., detrending the data)?
7. Physical Forcing
Q7a: You attribute the August 2009 PMOA enhancement to stronger winds (6.83 m/s median) and cooler SST. However, August 2009 also had higher micronekton biomass (11.2 gC/m² vs. 5.6 gC/m² in 2018). Could the PMOA difference be primarily biological rather than physical? A partial correlation or variance partitioning analysis would help separate biological vs. physical drivers.
Q7b: The wave Reynolds number is used as a proxy for bubble-mediated aerosol production. Did you test alternative parameterizations (e.g., whitecap fraction, bubble plume penetration depth) to see if the conclusions are robust? The Reynolds number depends on assumptions about seawater viscosity and salinity—were these varied in sensitivity tests?
7. Discussion & Interpretation
Q7a: The statement "diatoms, dinoflagellates, chlorophytes, and cryptophytes all covaried along with PMOA" (line 474-475) is supported by correlation analysis. But correlation does not establish causation. Could the PMOA-PFT correlations simply reflect shared seasonal forcing (e.g., light, temperature, nutrients) rather than direct biological emission? Please discuss how you rule out common environmental drivers.
Q7b: You invoke "viral lysis" as a mechanism for PMOA release (lines 507-508). Do you have any independent evidence for viral activity during the sampled periods? Without viral abundance data, this remains speculative. Consider tempering this interpretation or citing literature from the same region.
Q7c: The discussion of "homeoviscous adaptation" (line 399-401) linking colder SST to increased lipid unsaturation is interesting but speculative for your specific dataset. Did you measure lipid composition in seawater or aerosol? If not, this should be presented as a hypothesis rather than a conclusion.
Q7d: You note that "not all biological activity is represented (e.g., viruses, archaea, fungi)" (line 518). How significant are these omissions? Recent work suggests archaea and fungi can be important in sea spray aerosol formation. Please quantify or discuss the potential contribution of these missing groups.
Q7e: The conclusion states that "this study provides observational evidence that higher trophic levels contribute directly to PMOA" (line 499). Given that micronekton are not primary producers but consumers, is the correlation with micronekton simply tracking the same seasonal signal as phytoplankton? A partial correlation controlling for total PFT biomass would help establish whether micronekton add explanatory power beyond phytoplankton.
8. Technical/Editorial Issues
Q8.1: There are duplicate author affiliations and repeated text in the author list (lines 5-8 appear duplicated). Please correct.
Q8.2: The abbreviation "SPM" (suspended particulate matter) is introduced but not used in the main analysis. Was SPM considered? If not, remove from methods.
Q8.3: Several references in the text are incomplete or incorrectly formatted (e.g., "Daey and Wakeham_1986" in line 433, "Alberne et al., 2024" in references has no initials). Please check all references against the EGU reference style.
Q8.4: The manuscript uses both "micronekton" and "microenokton" (typo on line 165). Please standardize spelling throughout.
Q8.5: Figure numbers: Figure S5, S6 etc. are referenced but the supplement file was not provided in the review. Please ensure all supplementary figures are properly labeled and referenced.