Interactive comment on “ Long-term trends of global marine primary and secondary aerosol production during the recent global warming hiatus ( 2000 – 2015 ) ”

This manuscript aims to understand how marine derived aerosols changed during the warming hiatus from 2000-2015. The authors consider primary sea spray aerosol (SSA) and secondary aerosol formation from dimethylsulfide (DMS) fluxes. They compute the SSA fluxes using a parameterization based on sea surface temperature (SST) and wind speed (U10) from Gong (2003). DMS fluxes are computed using chlorophyll and mixed layer depth parameterizations for seawater concentrations from several publications and the Liss and Merlivat (1986) gas transfer parameterization. The aerosol optical depth (AOD) from both sources was computed using the model Optical Properties of Aerosols and Clouds (OPAC). In addition, the authors compare their computed marine derived AOD values with MODIS AOD values. They find that the annual global

1. Hiatus relevance -The authors state on lines 43-46 that the hiatus may be due to several factors including aerosols, but never really reach a conclusion about this.They refer to the ocean heat uptake and regional changes in wind later in the text to discuss changes in fluxes leading to changes in AOD, but never make a direct connection between the AOD and the hiatus.Did the aerosols contribute to the hiatus?This reads more like changes in physical conditions (e.g.U10) due to the hiatus caused changes in the precursor fluxes and the AOD.
ANS) The primary objective of this study is to analyze the long-term trends in global SSA emissions and DMS fluxes during 2000-2015 (i.e., corresponding the recent global warming hiatus period).The secondary and third objectives are to evaluate the long-term variations of separate aerosol loadings derived from primary (SSA) and secondary (DMS oxidation) marine aerosols and their contributions to satellite-observed AOD, respectively.Thus, we did not analyze causes of global warming hiatus because that is beyond the scope of this study.We changed the title of this paper, "Long-term trends of global marine primary and secondary aerosol production during 2000-2015."In addition, we removed two sentences in the "Introduction" to clarify our objectives.However, as the reviewer suggested, we further analyzed the correlation between aerosol properties (AOD) and physical conditions (SST and SSW) to find their connection during 2000-2015 (see Fig. 4 in the revised text and Fig. 1).For example, we analyzed the relation of SST with atmospheric aerosol loadings (MODIS AOD), as well as other parameters (e.g., SSW and chlorophyll).New discussion on this is imbedded in the revised manuscript.Fig. 1.Distribution of correlation of (a) SST vs. U10, (b) SST vs. Chl-a, (c) SST vs. MODIS AOD, (d) U10 vs. MODIS AOD, (e) SST vs. AODSSA, (f) SST vs. AODDMS, (g) U10 vs. AODDMS, and h) U10 vs. Chl-a.
Note that the 1951-2012 period represents 61 years of climate data, while the 1998-2012 period represents only 15 years; a fifteen-year period is well known to be too short to reliably detect long-term climate trends in global mean surface air temperature.The slowdown in warming during this fifteen-year period, often termed a "hiatus" in global warming can be ascribed, either fully or in part, to natural variability in the climate system (Roberts et al., Nature, 2015; https://www.nature.com/articles/nclimate2531). Discussion on this is added to the middle part of page 9.In addition, we replaced U10 and SST data obtained from two satellites (QuikSCAT and ASCAT satellites) with ECMWF reanalysis data due to limitation of discontinuity of two satellite measurements for the trend analysis (see Supplementary Fig. 1).Relevant figures and tables were revised.
2. From lines 74 to 80, the authors state that there are many unknowns related to aerosol production/loading and claim that, therefore, their goal is to study aerosol C3 trends during the recent hiatus.However, I am not sure one has anything to do with the other.They could study aerosol changes over any period to investigate these uncertainties.
ANS) The sentences from lines 74 to 80 means that "the studies of the long-term variation in marine aerosol production, as well as its contributions to aerosol concentration over the ocean, are scarce", compared to the studies related to aerosol radiative forcing by changes in anthropogenic aerosol emissions.We revised the sentence to avoid confusion (in the upper part of page 4).Note that there is significant discrepancy in the magnitudes of SSA emission rates and DMS fluxes due to the difference in parameterization methods for the flux estimation.Our main objective was not to estimate these fluxes, but to identify the trend of temporal variation of SSA emission and DMS flux during 2000-2015 using the most widely used method, Gong (2003).Although there is large uncertainty in the magnitude of SSA emission rates, the estimation of long-term trend (e.g., upward, downward, on no change) is less uncertain because of the one parameterization method usage.

Specific comments:
3. Lines 54-57: Are there any other studies the author can cite to corroborate the findings cited (i.e.Klimot et al., 2017)?Also, how important are other sources of marine aerosol not included here (e.g.glyoxal, isoprene)?ANS) Additional references (Street, et at., 2009;Jaeglé, et al., 2011;de Leeuw, et al., 2011;Ovadnevaite, et al., 2014;Janssens-Maenhout et al., 2015) were added to the revised manuscript (in the lower part of page 3).In addition, some discussions on other sources of marine aerosol (e.g., glyoxal, isoprene) were newly added to the revised text (in the middle part of page 3)."In contrast to the importance of isoprene as a biogenic secondary organic aerosol (SOA) precursor in the continents, Arnold et al. ( 2009) suggested an insignificant role of isoprene in the remote marine SOA formation.SOA production from the oxidation of marine VOCs (dialkyl amine salts, marine hydrocarbon, glyoxal, etc.) was found to be insignificant, compared with natural SSA emissions (Fu et al., 2008;Myriokefalitakis et al., 2008;Myriokefalitakis, et al., 2010;Rinaldi et al., 2011;Mahajan et al., 2014)."natural, and anthropogenic emission sources also indirectly reflected ambient aerosol changes during the study period.Discussion on this was added to the "Introduction." Reference: Li, J., Carlson, B.E., Lacis, A.A.: How well do satellite AOD observations represent the spatial and temporal variability of PM2.5 concentration for the United States?Atmos.Environ.102, 260-273.2015.6. Lines 122-123: It is very well known that DMS concentrations in seawater are hard to predict and not really correlated with chlorophyll.The values depend highly on the presence of DMSP producing phytoplankton and DMSP cleaving phytoplankton and bacteria.The parameterizations used have some value in certain areas, sometime, but may not be correct all of the time.Did the authors do any sensitivity tests to compare different parameterization methods before settling on this set of equations?ANS) As the reviewer commented, DMS concentrations in seawater are not perfectly correlated with chlorophyll-a and the parametrization used in certain areas, sometime, may not be correct.The correlation between DMS and chlorophyll-a was re-examined using the updated monthly global DMS climatology dataset by Lana et al. (2012).Lana et al. (2012) showed the good correlation (r2 = 0.47) between DMS and chlorophyll-a concentrations and estimated the seawater DMS uncertainty of −25%/+15% (see Fig. 2).The current available data to estimate the global distribution of DMS concentration in seawater during a long-term period (16 years) are global chlorophyll-a data from the European Space Agency (ESA) GlobColor (www.globcolour.info/).We also performed the sensitivity test of chlorophyll-a to seawater DMS.25% change in the chlorophyll-a concentration resulted in +1/-2% change in DMS.The related sentences were added to the upper parts of pages 6 and 13.  )."According to Elliott (2009), there is also some indication that the lower end values of kw of Liss and Merlivat (1986) may ultimately be preferred."The above sentences were added to the middle-bottom parts of page 6.Thus, our DMS flux estimation is not likely to be significantly underestimated.
Upstill-Goddard, R. C.: In situ evaluation of air-sea gas exchange parameterizations using novel conservative and volatile tracers, Global.Biogeochem.Cy., 14,373-387, 2000.2017)?I realize the timing of the two studies may not be identical, but Quinn et al. take a global view and discuss the controls on SSA and DMS derived aerosol.Also, evidence from direct DMS flux studies suggests that seawater concentrations are more important that U10 for fluxes.ANS) In this study, we analyzed SSA mass emission rates.However, Quinn et al. (2017) estimated the contribution of SSA to the number concentration of CCN (and total particle number concentrations) in the marine boundary layer using a lognormal-modefitting procedure (Aitken, accumulation, and SSA modes).Thus, direct comparison of our result with Quinn et al. ( 2017) is not reasonable because the number fraction of SSA depends on the Aiken mode, whereas the mass fraction of SSA on a coarse mode.We incorporated finding of Quinn et.al (2017) in the revised manuscript (in the upper part of page 10)."However, SSA emissions makes a contribution of less than 30% to CCN concentrations increasing with wind speed, on a global basis, with the exception of the high southern latitudes (Quinn et al., 2017)."As the reviewer commented, in the biologically productive areas, seawater DMS concentration can be more important than wind speed for sea-to-air DMS flux.However, since most regions of the world's oceans are oligotrophic, i.e., they contain very low levels of nutrients, DMS concentrations in seawater are low (see Fig.  (1986) and Wanninkhof (1992) parameterizations for comparison.The annual DMS fluxes were also estimated using monthly data with a resolution of 1 • × 1 • between -90 • and 90 • over the global ocean.Meanwhile, in our study, the global DMS fluxes were estimated using both surface seawater DMS concentrations estimated from a DMS empirical algorithm and the gas transfer velocity derived from the Liss and Merlivat (1986), including the monthly data with 25-km spatial resolution between -60 • and 60 • .Because there were somewhat differences (e.g., DMS measurement and covering area (km2)) between our and previous studies, we indirectly compared with the previous studies.Discussion on this was added to the lower part of page 11. "The flux estimates using the mean flux derived from the region of 60 11.Lines 283-284: I am not sure I follow the "thus" logic; does the comparison of the trends before this sentence make sense in another way?ANS) As the reviewer suggested, we removed these sentences.
12. Line 290: What is the adjustment factor?
ANS) The adjustment factor (ï Ąś) in Eq. ( 1) is an adjustable shape parameter that controls the submicron size distribution.The description of the adjustment factor is given in the section 2.1.ANS) Conclusion was revised as the reviewer suggested, including short discussion on the global warming hiatus.17.Tables 1 and 2: Need better description of units ANS) More description of units was added.18. References Quinn, P. K., Coffman, D. J., Johnson, J. E., Upchurch, L. M., Bates, T. S. (2017) Small fraction of marine cloud condensation nuclei made up of sea spray aerosol, Nature Geoscience (10), http://dx.doi.org/10.1038/ngeo3003.ANS) As the reviewer suggested, the reference (Quinn et al., 2017) was cited in the C13 revised text.

General comments:
After closer reading of this manuscript, I am unable to recommend this manuscript for publication in ACP.The subject of the manuscript is a calculation of inferred/estimated trends in DMS and sea spray aerosol emissions over the years 2000-2015, as well as trends in the AOD contributed by these aerosol sources, which may have occurred as a consequence of variability in wind speed and sea surface temperature during this time period.However, the methods used have such serious deficiencies as to make the results uninterpretable.I was surprised to find no discussion or evaluation of potential instrument biases, discrepancies, retrieval uncertainties, or long-term drifts in satellite measurements here.Without this information, it is impossible to determine whether any trends in the wind speed or SST time series used here are likely to be reliable reflections of real-world trends.
In particular, in the analysis performed here, for both wind speed and sea surface temperature, two different satellites were used with breaks in the long-term time series The first reviewer has already pointed out some of the (well-known) challenges in estimating oceanic DMS concentrations.Some comparison with observed ocean DMS concentrations and fluxes would be useful in evaluating how well this approach produces results that resemble the real world.The calculation of AOD associated with marine aerosol sources, as described in the manuscript, is also overly simplistic and will not produce interpretable results: -L.
ANS) As the reviewer commented, DMS concentrations in seawater are not perfectly correlated with chlorophyll-a and the parametrization used in certain areas, sometime, may not be correct.The correlation between DMS and chlorophyll-a was re-examined  2).According to Elliott (2009), there is also some indication that the lower end values of kw of Liss and Merlivat (1986) may ultimately be preferred.Therefore, in this study, the Liss and Merlivat (1986) approach was used to calculate the DMS fluxes during the study period."These sentences were added to the parts of pages 6 and 7. Thus, our DMS flux estimation is not likely to be significantly underestimated (see Fig. 3 and 4).In light of such analyses, this "hiatus" period is more realistically viewed as a part of the expected natural variability in the climate system.
ANS) The main goal of this study is to examine whether there is a long-term trends of global marine primary and secondary aerosol production during 2000-2015 or not.We do not focus on the reasons of recent hiatus (1998-2012).As the reviewer suggested, we revised the text (adding a following paragraph)."The recent hiatus was attributed to some combination of external climatic forcing that are not represented adequately in the model simulation and the internal (natural

Fig. 3 .
Fig. 3.Table for summary of environemntal conditions in air and water during gradient flux measurements.Table for Transfer Velocityies Applied in the Present work.

Fig. 4 .
Fig. 4. Comparison of transfer velocity parameterization schemes with observations.8. Paragraph starting at 189: How do the findings presented in this paragraph compare with the findings in Quinn et al. (2017)?I realize the timing of the two studies may not be identical, but Quinn et al. take a global view and discuss the controls on SSA and DMS derived aerosol.Also, evidence from direct DMS flux studies suggests that seawater concentrations are more important that U10 for fluxes.

Fig. 2 .
Fig. 2. Linear regression of DMS concentration against Chl-a."Recently, Smith et al. (2018) measured directly the DMS fluxes made using two independent methods: the eddy covariance (EC) technique and the gradient flux (GF) technique during Feb-Mar 2012 in the southwest Pacific Ocean.Their results showed good agreement between these two independent methods. .The two parametrization methods (kw) of Liss and Merlivat (1986) and Wanninkhof (1992) were evaluated using flux measurement data set reported by Smith et al. (2018).The evaluation of the gas transfer velocity showed the lowest root mean square error (RMSE) between the kw calculated from Liss and Merlivat (1986) and indirect measurements of kw by Smith et al. (2018) using the mean wind speeds, DMS concentrations, and flux measurements (Supplementary Fig.2).According to Elliott (2009), there is also some indication that the lower end values of kw of Liss and Merlivat (1986) may ultimately be preferred.Therefore, in this study, the Liss and Merlivat (1986) approach was used to calculate the DMS fluxes during the study period."These sentences were added to the parts of pages 6 and 7. Thus, our DMS flux estimation is not likely to be significantly underestimated (see Fig.3 and 4).

Fig. 3.
Fig. 3.Table for summary of environemntal conditions in air and water during gradient flux measurements.Table for Transfer Velocityies Applied in the Present work.
Table for summary of environemntal conditions in air and water during gradient flux measurements.Table for Transfer Velocityies Applied in the Present work.
satellite-buoy observations during their period of overlap (November 2008 through November 2009).An examination of the collocated winds shows a high degree of agreement in direction, but reveals systematic differences in wind speed that depend on rain rate, the strength of the wind, and SST.Time mean difference (ï Ą ĎW=WQS-WAS) between collocated QS and AS 10 m neutral wind speed estimates is generally less than 1 m/s (Bentamy et al. (2012), Figs.3 and 4).After applying the correction function, 0.18 m/s positive bias in ï Ą ĎW was estimated (Bentamy et al. (2012), Figure9and Guo et al. (2018), Table2).The biases of QS and AS between buoy-derived mean wind speeds and satellite-derived mean wind speeds at a height of 10 m above sea level are 0.23 and 0.09 m/s, respectively(Guo et al., 2018).The slopes between buoy-derived mean wind speeds vs QuikSCAT and that vs ASCAT are 1.03 and 1.01, respectively (Guo et al., 2018, Table2).Thus, we compared European Centre for Medium-Range Weather Forecasts (ECMWF, www.ecmwf.int)reanalysisdatawith QukSCAT and ASCAT wind speeds (see Fig.7).For accurate trend analysis with observation consistency, we reanalyzed entire trend analysis using monthly U10 and SST data at 25-km spatial resolution obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF, www.ecmwf.int)for 2000-2015.
different periods?Since the analysis focusses on time series, this information is critical to understanding the causes of the inferred trends, and ruling out potential instrument bias.ANS) Comparisons to independent mooring and shipboard observations byBentamy et al. (2008), Verspeek et al. (2010), Guo et al. (2018) show that ASCAT wind speed has accuracies similar to QuikSCAT.Furthermore, Bentamy et al. (2012) examined a set of space-time collocated observations for QuikSCAT (QS) and ASCAT (AS) and triply collocated satellite- Table for summary of environemntal conditions in air and water during gradient flux measurements.Table for Transfer Velocityies Applied in the Present work.Fig.4.Comparison of transfer velocity parameterization schemes with observations.The AOD calculation procedure is not simplistic, which use the aerosol size distribution with lognormal-mode-fitting procedure, microphysical and optical properties of aerosols, etc. Description on this is given in the upper part of page 8 and detailed information the OPAC model is given Hess et al. (1998).It's not possible to realistically calculate the AOD from emitted aerosol without treatment of its transport, microphysical evolution, and removal in the atmosphere.The distribution of SSA concentrations in the atmosphere is controlled by removal processes (e.g. the distribution of precipitation) at least as much as by emission processes.ANS)The OPAC model used in this study needs aerosol number concentration to simulate the AOD.The aerosol number concentration is affected by emission, transport, heterogeneous reactions, microphysical evolution, removal in the air, etc.In this study, It's not possible to generate reasonable distributions of DMS derived sulfate aerosol in this way, with no consideration of atmospheric transport and chemistry (reaction with OH and subsequent reaction mechanisms), which occurs over time scales of several days to weeks.There is a large literature on the production of sulfate aerosol from DMS-derived SO2, which shows the critical role of transport and chemistry processes in determining the contribution of DMS-derived SO2 to global aerosol concentrations.As just one example of this, in a paper that is referenced by the authors, Korhonen et al.(2008), Figure8clearly shows that the geographic distributions DMS emissions, and of the increase in cloud condensation nuclei number associated with DMS emissions, are nearly uncorrelated.These are major flaws in the methodology that need to be addressed before any of the results will be interpretable.strongercontrolonclimate(throughCCNconcentrationchange)than net increases in biological productivity.Seasonal covariance of aerosol and methane sulfonic acid (MSA), an aerosol phase oxidation product of dimethyl sulfide (DMS), was also used by Ayers and Gras (1991) and Ayers et al. (1991) as evidence supporting a major role of a marine biogenic reduced-sulfur source in driving the CCN seasonal concentration cycle.Thus, open questions remain on the source of MBL NSS-sulfate with possibilities including DMS oxidation in the MBL that leads to particle growth to the accumulation-mode size range through vapor condensation and accumulation of mass during cloud processing(Hegg et al., 1992; Hoppel et al., 1994).Note that our results are not number concentration, but AOD which indirectly semi-quantify the aerosol mass loading.As mentioned before, due to the limitation of computing time for 16 years simulation over the global oceans, we used our best approach to estimate the AODDMS based on DMS flux and sulfate aerosol number concentration.Our results showed similar result to Gondwe et al.(2003).For example, our mean contribution of DMS (∼50%) to AOD in the Southern Hemisphere were similar to mean annual contribution of DMS to the climate-relevant non-sea-salt sulfate column burden (43%) by Gondwe et al.(2003).Our mean contribution in the Northern Hemisphere were less (8% during MAM and JJA), where anthropogenic sulfur sources are overwhelming.Gondwe et al.(2003)reported 9% in the Northern Hemisphere.In a previous study, the contribution of biogenic sulfur (DMS) in fine particle mode over the Atlantic Ocean was less than 35% of the excess sulfur in the NH (0ï Ćř-60ï ĆřN), and approximately 60% in the SH (0ï Ćř-35ï ĆřS, Patris et al., 2000).Our mean DMS-derived sulfate contributions to atmospheric aerosol loading were similar to these values, 30% in the NH and 50% in the SH (51% for 0ï Ćř-30ï ĆřS).This suggested that out approach was not unrealistic.climate warming at a rate of 0.2 deg C per year, periods of cooling occur frequently.For example, a -0.10 deg/yr cooling period lasting fifteen years occurs more than 10% of the time (Roberts et al., Nature, 2015; https://www.nature.com/articles/nclimate2531).
the NH, where anthropogenic sulfur sources are overwhelming.A recent study by Quinn et al. (2017) found that Aitken-mode particles make up a large fraction of the CCN in the Southern Hemisphere at high supersaturations (>0.5%).Note that the timescale for oxidation of DMS is on the order of hours to one day.Longer temporal scale might not observe correlation of monthly data with CCN number concentration.Woodhouse et al. (2003) suggested that changes in the spatial distribution of DMS emissions (through changes in the phytoplankton population or wind speed patterns) could exert a climate variation (IPCC 2013 and references therein).Roberts et al. (2015) estimated an approximate 10% probability of a 10-year warming hiatus due to internal variability given an expected contemporary warming rate of approximately 0.2ï ĆřK per decade." 5. L. 43-44: This sentence should be updated: "According to a recent study (England et al., 2014), the global mean SAT has remained flat since around 2001.".It is now 2018; the global mean surface air temperature has clearly risen between 2001 and 2018.ANS) The sentence was removed.Interactive comment on Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-322,2018.