First observations of aerosol mass fluxes over open water using a large-aperture scintillometer

Liu, Renjie; Pu, Yuhong; Yuan, Renmin; Hang, Chaoxun

doi:10.5194/egusphere-2026-2182

Preprints

https://doi.org/10.5194/egusphere-2026-2182

Preprints

19 May 2026

| 19 May 2026

First observations of aerosol mass fluxes over open water using a large-aperture scintillometer

Renjie Liu, Yuhong Pu, Renmin Yuan, and Chaoxun Hang

Abstract. Near-surface aerosol fluxes over open water are critical for understanding air–sea interactions and atmospheric radiative budgets, yet direct micrometeorological observations remain very limited. To address this gap, this study presents the first application of a large-aperture scintillometer (LAS) for retrieving aerosol mass fluxes over open-water environments, including an inland lake and a coastal bay. LAS-derived fluxes were compared with conventional eddy-covariance (EC) measurements and independent dry-deposition calculations. Under conditions without severe optical attenuation, the LAS-derived fluxes exhibited temporal variations broadly consistent with EC measurements, with normalized root-mean-square errors (NRMSEs) to 15.0%–16.8%. The coastal campaigns revealed the operational boundary of this optical approach: dense fog and persistent high humidity severely attenuated the LAS signal and caused extended data gaps. Comparisons with dry-deposition calculations further showed that the path-averaged LAS measurements may better capture site-scale variability at the coastal site. Overall, these results demonstrate the potential of LAS as a valuable complements to traditional point measurements for characterizing aerosol exchange over open water.

Received: 16 Apr 2026 – Discussion started: 19 May 2026

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

Download & links

Preprint (PDF, 2805 KB)

Supplement (2807 KB)

Download & links

Renjie Liu, Yuhong Pu, Renmin Yuan, and Chaoxun Hang

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-2182', Anonymous Referee #1, 01 Jul 2026
The paper addresses a relevant question: Can a large-aperture scintillometer be used for accurate measurements of aerosol mass fluxes over open water? The experimental set-up of using EC and dry deposition filters seems appropriate to answer this question. However, in the way the method is executed and the data is analyzed, quite some issues come forward that must be addressed.
The dry deposition filters for measuring the aerosol mass, used for determining the aerosol fluxes, are handled in ways that do not seem to comply with European or American standards. See https://www.nature.com/articles/s41598-026-42411-4 for a paper that explicitly mentions these standards. The authors should indicate what kind of standard they follow with their method and how this method or standard relates to the European or American standards for measuring particulate matter. Of particular concern is the condensation of gaseous organic carbons, see https://acp.copernicus.org/articles/9/7243/2009/.

The use of 30 min intervals, 2 hour intervals, (moving) averages, and outlier rejection greatly impacts the statistical analyses used by the authors. Stating that the correlation for a 30 min interval is low and that the deviation between 2hr averages is also low, seems more like an application of the Central Limit Theorem than an actual physical result. This also impacts the ensemble averages, but there the averages become more problematic due to the large spread (2 orders of magnitude) of the aerosol fluxes. In those cases the average of the logarithmic values (the geometric mean) or the use of the median would be more appropriate. Since the authors already calculate and display the median, they would do well to redo their analysis and conclusions based on the median values for the ensemble. Regarding the selection bias, the authors clearly show that removing data due to fog and rain impacts the comparison. Instead of seeing this as a reason to do follow-up analysis where the exact same selection was applied to the measurements, they weaken their conclusions and caution the reader for doing this. The authors should apply identical selections to the measurements in order to attain results from which stronger conclusions can be attained.

Lastly, the authors have done a comparison between LAS and EC before, but on land. The performance analysis of LAS versus EC at coastal environment should be compared with the same comparison for land environment. This is necessary to show the limitations of LAS, for which they now only give vague qualitative answers but no quantitative answers such as thresholds (for example the amount of attenuation by fog/particles where the LAS starts to give deviating answers, and thus data should be excluded) that allow reproduction by fellow scientists. Moreover, this manuscript contains the theory associated with land, while only fleetingly mentioning coastal influences.

Next to these issues there are several points where the quality of the paper or results would improve. The authors are suggested to follow-up on these points.
Figures 2 and 5 show transparent data points for the 30 min averages, while the 2 hr block values are shown non-transparent. The current transparency is so high that the 30 min average values are not seen as they are equally as bright as text on the other side of the paper seen through the paper. Suggestion: less transparent points or non-transparent points in other colors.

Exclude the barbecue smoke data points

Explain in the appendix better why the choice should be 30 min averages and not 1 hr averages, as the spectral behavior is minimal at 1 hr frequency.

Explain why a gas analyzer was not associated with the EC system, and how its absence impacts the measurements.

Explain what kind of behaviors characteristic for coastal environments you expect to see with these measurements, and why these effects should be better or worse visible with an LAS system.

Explain why the chosen set-up is chosen and not other set-ups like aerosol mass flux measurement instead of aerosol dry-deposition filtering method.

Explain why the 0.1 mg precision of the analytical balance is sufficient to attain 0.0001 ug precision in the aerosol flux.

Be explicit about point-average and path-average for LAS at heterogeneous surfaces.

Add a table of equipment used, similar to table 1 for meteorological conditions.

Consider adding relevant references to statements made in the introduction that are now lacking references.

Add information regarding the calibration of the different sensors associated with the sonic anemometer, as they have different sampling frequencies, and thus could sample different volumes of air.

Explain in more detail how the wind directions were selected to reduce flow distortion.

Explain to what extent PM10 is a good approximation or proxy for aerosol mass concentration.

Indicate how many data points are used in the ensemble.
Citation: https://doi.org/10.5194/egusphere-2026-2182-RC1
RC2:
'Comment on egusphere-2026-2182', Anonymous Referee #2, 05 Aug 2026
General assessment
The manuscript investigates whether a large-aperture scintillometer (LAS) can be used to estimate aerosol mass fluxes above water surfaces. Measurements above an urban lake and a coastal bay are compared with an extinction-based eddy-covariance (EC) approach and filter-based dry-deposition estimates. The subject is relevant to Atmospheric Measurement Techniques, and spatially averaged aerosol-transport measurements using scintillometry could be valuable.
However, the manuscript does not yet provide sufficient methodological information or evidence to demonstrate that the reported quantities are quantitatively reliable aerosol mass fluxes. The main concerns involve the response of the visibility sensor, the absence of spectral flux diagnostics, the sign and physical meaning of the LAS-derived quantity, the applicability of the free-convection formulation, hygroscopic effects on the extinction-to-mass conversion, filter treatment, and the interpretation of modelled dry deposition as an independent validation method.
I therefore cannot recommend publication in its present form. Major revision and substantial reanalysis are required. In particular, the LAS–EC validation can be retained only if the authors demonstrate that the visibility sensor provided independent extinction measurements with sufficient temporal response to resolve turbulent covariance.
Specific comments
1. Flux cospectra and cumulative ogives should be calculated for all accepted averaging intervals and all four campaigns. Velocity spectra alone cannot demonstrate that the covariance between vertical wind and aerosol extinction represents a resolved turbulent flux. Cospectra are required to identify the frequencies contributing to the flux and diagnose high-frequency attenuation, noise, time-lag errors, aliasing and phase shifts. Ogives should show whether the covariance converges within the 30 min averaging period or remains affected by low-frequency non-stationarity and mesoscale variability. Campaign-level median cospectra and ogives with interquartile ranges could be presented in the manuscript, with individual diagnostics supplied in the supplement or data repository. Equivalent LAS log-intensity spectra are also needed to support the separation of the high- and low-frequency components used to derive the real and imaginary refractive-index structure parameters.
The manuscript must distinguish between logging frequency, output interval, temporal resolution and response time (check, e.g., Buzourius et al., 2003). Although PWD50 data were reportedly recorded at 1 Hz, the standard instrument specification gives a measurement interval of 15 s. Polling the output every second may therefore only reproduce internally averaged values rather than provide independent 1 Hz extinction measurements. This corresponds to a maximum independent Nyquist frequency of approximately 0.033 Hz, even before additional response attenuation is considered. A longer averaging period cannot recover covariance already removed by the instrument response.
The authors should report the sensor’s independent update interval, internal averaging, physical response function, synchronization with the sonic anemometer, time-lag procedure and frequency-response correction. They should also clarify whether any hardware or firmware modification was implemented and experimentally characterized. This is particularly important because Yuan et al. (2024) used a modified CS120A sensor and still reported response-time and high-frequency-noise limitations. If a standard PWD50 output was used, the resulting estimate should be described as a low-frequency covariance proxy rather than a conventional 1 Hz aerosol EC flux.
2. It is unclear whether Eq. (5) represents turbulent transport, flux magnitude, net flux, total flux, surface emission or another effective flux. Because the equation contains only positive powers of structure parameters, it appears to return a magnitude without determining direction. Nevertheless, all results are positive and repeatedly interpreted as upward aerosol emission, while logarithmic axes prevent negative deposition fluxes from being displayed.
Continuous upward emission during all campaigns is physically unlikely, particularly under weak winds or stable conditions, when downward deposition should occur. The authors must state explicitly whether absolute flux magnitudes are reported and explain how flux direction is determined.
Equation (5) appears to be a free-convection approximation whose validity is mainly expected under strongly unstable conditions. Its application to daytime and nighttime periods, different seasons and a mechanically turbulent coastal site is not justified. The authors should either use a complete Monin–Obukhov similarity formulation or restrict the analysis to conditions satisfying the free-convection assumptions. A stability-stratified comparison based on (z/L), friction velocity and the air–water temperature difference is essential. The assumption that solar radiation automatically generates unstable daytime conditions above water is insufficient because the thermal inertia of the water surface may produce a different diurnal stability cycle.
3. The term “open water” is insufficiently defined and may be misleading. An urban campus lake and a sheltered coastal bay do not necessarily represent open-ocean or high-fetch conditions. The authors should clarify whether the term refers only to an unobstructed water surface and quantify fetch, surface homogeneity and the fraction of each footprint located over water. If substantial land contributions occur, the results should be described as exchange over heterogeneous lake–land or coast–water systems.
All inputs and validity criteria for the Kljun et al. (2015) model, the wind-coordinate transformation and the treatment of boundary-layer height should be reported. In addition to campaign-mean footprint climatologies, the authors should calculate an instantaneous water-footprint fraction and LAS–EC footprint-overlap metric. Otherwise, footprint mismatch can only be proposed as a possible explanation for the low correlations, not presented as a demonstrated cause.
4. Section 2.2 does not provide enough information to reproduce the EC calculation. The authors should specify the coordinate-rotation and detrending methods, time-lag determination, resampling and synchronization, spike detection, stationarity and integral-turbulence tests, low u* filtering, data-completeness criteria, flow-distortion treatment and high- and low-frequency spectral corrections. They should also discuss air-density and water-vapour effects, whether a Webb–Pearman–Leuning-type correction is applicable to the converted mass concentration, and how hygroscopic effects influence the covariance. Random uncertainty, instrumental detection limits and uncertainty in converting extinction into mass should be quantified. Zinke et al. (2024) may provide a useful methodological reference.
The ±3 standard-deviation filter relative to a 2 h moving average is insufficiently described. If applied to raw high-frequency variables, it could remove genuine turbulent excursions and attenuate covariance. Despiking of raw signals should be clearly distinguished from quality screening of averaged fluxes.
5. The conversion factor R_MN is central to both LAS and EC estimates but is not adequately determined. Filter measurements represent conditioned aerosol mass, whereas LAS and the visibility sensor respond to particles at ambient relative humidity, which reached 100%. Hygroscopic growth can strongly increase particle size and extinction near saturation, while fog droplets may contribute to extinction without representing the PM₁₀ mass collected on filters.
The authors should therefore evaluate the humidity dependence of extinction, scattering enhancement, hygroscopic growth and possible humidity–aerosol covariance rather than assume a constant R_MN. The possible contribution of water-vapour fluctuations to the real refractive-index structure parameter should also be discussed. Finally, the imaginary component should not be described as responding only to light-absorbing aerosol: the measured attenuation includes scattering, which is generally important or dominant for marine aerosol.
6. The filter-conditioning procedure requires justification. Conditioning glass-fibre filters for only 3 h at 45 °C without reported relative humidity does not conform to established ambient-PM reference procedures. EN 12341:2023 requires conditioning at 19–21 °C and 45–50% RH for at least 48 h and a balance resolution of no more than 10 µg. The relevant US EPA PM₁₀ and PM_2.5 methods require at least 24 h of controlled temperature and humidity conditioning; the PM_2.5 method additionally specifies PTFE filters and microgram-level weighing. Heating loaded filters to 45 °C may remove ammonium nitrate, semi-volatile organic matter and particle-associated water, biasing the measured mass.
The authors should identify the standard or validated protocol followed and report the filter manufacturer, material and dimensions; sampling duration and sample numbers; whether PM_2.5 and PM₁₀ were sampled simultaneously; flow calibration; transport and storage conditions; field and laboratory blanks; repeated weighings and mass-stability criteria; detection limits; and measurement uncertainty. If PM_{2.5 -10}was obtained by subtracting parallel PM_2.5 from PM₁₀ measurements, this must be stated and the uncertainty propagated.
The deposition velocity calculation is not reproducible. Explicit equations are needed for aerodynamic and surface resistances, Brownian diffusion, inertial impaction and settling velocity because the cited formulations are not identical. All assumptions should be reported, including particle size distribution, density, shape factor, slip correction, hygroscopic growth, surface roughness, friction velocity, stability treatment and water-surface collection efficiencies. The numerical deposition velocities used for each fraction must also be given. Assigning one representative velocity to PM_2.5 and another to PM_2.5_{– 10}may produce large errors because deposition velocity varies by orders of magnitude within these broad size ranges.
The filter-based calculation combines measured concentration with parameterized deposition velocity and is therefore a model estimate rather than an independent direct flux measurement. It should not be used as decisive validation of the LAS method. The authors must also show that the stated balance precision of 0.1 mg supports deposition fluxes reported to four decimal places after propagation of mass, volume and model uncertainties.
7. The reported meteorological averages are insufficient to assess whether the flux-method assumptions were satisfied. Mean wind speeds of approximately 0.96–1.19 m s^-1 indicate that weak-turbulence periods may have been common during measurements. Campaign distributions should be provided for wind speed and its variability, friction velocity, standard deviations, turbulent kinetic energy, z/L, stationarity, water temperature, air–water temperature difference and boundary-layer height. Wave height or another sea-state indicator might be also useful at the coastal site. Because boundary-layer height is an input to the footprint model, its source and associated sensitivity should be documented.
The abstract and conclusions describe the method as operationally useful without specifying its performance envelope. The authors should provide a table defining the ranges of wind speed, turbulence, stability, relative humidity, extinction and received signal for which valid retrievals were obtained, together with data availability and rejection rates. Fog should be defined using a reproducible signal or extinction threshold rather than treated qualitatively.
8. Validation statistics must be calculated from strictly paired LAS and EC observations. If EC measurements continued while LAS data were unavailable during fog, the unpaired periods should not affect comparisons of campaign means, MBE, RMSE or NRMSE. They may instead be used to quantify differences in operational availability.
The exact NRMSE formula and normalization, sample number, median, standard deviation, quartiles, range and confidence intervals should be reported for every comparison. Because fluxes span approximately two orders of magnitude, their distributions should be assessed before using arithmetic means; median, geometric-mean or logarithmic analyses may be more appropriate. Paired statistical tests, regression slopes and intercepts with uncertainty, and methods accounting for error in both variables, such as Deming or orthogonal regression, should be considered. Bland–Altman analysis, a concordance metric and treatment of temporal autocorrelation would provide a more informative method comparison. Campaigns should first be assessed separately before pooling.
Improved agreement after 2 h averaging may result from smoothing and does not demonstrate that averaging resolves footprint differences. The 30 min data should remain the primary validation, with 2 h results presented as a sensitivity analysis. Similarly, the difference between correlations of (r=0.442) and (r=0.346) does not establish superior LAS performance without confidence intervals and a test for dependent correlations. Barbecue-smoke periods should be handled using objective exclusion criteria or a predefined sensitivity analysis rather than only a post hoc interpretation based on field notes.
9. Daytime maxima and coastal events are repeatedly attributed to sea-spray emission or bubble bursting, but no aerosol composition, size distribution, sea-salt tracer, wave-state or whitecap data are presented. The urban lake and coastal bay may also be influenced by continental and local anthropogenic aerosol.
If sea-spray emission is proposed, fluxes should be analysed against wind speed while controlling for stability and water-footprint fraction. Separate upward and downward EC fluxes and PM_2.5/PM₁₀ partitioning would also be informative. Without compositionally or size-resolved evidence, the observed optical-flux variability should not be attributed specifically to sea spray.
10. “Available upon request” is inadequate for a methodological validation study whose conclusions depend strongly on signal processing, quality control, averaging and data selection. In accordance with FAIR principles, used data should be deposited in a repository with a persistent identifier. If public release is not possible during review, the authors should provide reviewer access and commit explicitly to publication of the data and code upon acceptance.
11. The manuscript contains many minor typos and glitches. I suggest a specialised English proofreader should check the text. I'm not qualified to check the language.
12. Reference problems
“Seinfeld and Pandis (2016)” is cited at line 154 but absent from the reference list.

Wilczak et al. (2001) concerns sonic-anemometer tilt correction and does not support the aerosol covariance equation as cited.

Yang et al. (2016) concerns sulfur dioxide attribution, not aerosol-particle EC fluxes.

Kiehl and Briegleb (1993) is not an appropriate source for the statement that oceans cover 71% of Earth and are a major aerosol source.

Twomey (1977) supports aerosol–cloud albedo effects, not the direct scattering/absorption statement as currently written.

References
Buzorius, G., Rannik, Ü., Nilsson, E. D., Vesala, T., & Kulmala, M. (2003). Analysis of measurement techniques to determine dry deposition velocities of aerosol particles with diameters less than 100 nm. Journal of Aerosol Science, 34(6), 747-764.
Kljun, N., Calanca, P., Rotach, M. W., & Schmid, H. P. (2015). A simple two-dimensional parameterisation for Flux Footprint Prediction (FFP). Geoscientific Model Development, 8(11), 3695-3713.
Yuan, R., Zhang, H., Hua, J., Liu, H., Wu, P., Zhu, X., & Sun, J. (2024). Comparison of the imaginary parts of the atmospheric refractive index structure parameter and aerosol flux based on different measurement methods. Atmospheric Measurement Techniques, 17(7), 2089-2102.
Zinke, J., Nilsson, E. D., Markuszewski, P., Zieger, P., Mårtensson, E. M., Rutgersson, A., ... & Salter, M. E. (2024). Sea spray emissions from the Baltic Sea: comparison of aerosol eddy covariance fluxes and chamber-simulated sea spray emissions. Atmospheric Chemistry and Physics, 24(3), 1895-1918.
Citation: https://doi.org/10.5194/egusphere-2026-2182-RC2

Renjie Liu, Yuhong Pu, Renmin Yuan, and Chaoxun Hang

Supplement

https://doi.org/10.5194/egusphere-2026-2182-supplement

Renjie Liu, Yuhong Pu, Renmin Yuan, and Chaoxun Hang

Viewed

Total article views: 276 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
197	62	17	276	36	13	13

HTML: 197
PDF: 62
XML: 17
Total: 276
Supplement: 36
BibTeX: 13
EndNote: 13

Views and downloads (calculated since 19 May 2026)

Month	HTML	PDF	XML	Total
May 2026	133	34	8	175
Jun 2026	28	7	4	39
Jul 2026	23	17	3	43
Aug 2026	13	4	2	19

Cumulative views and downloads (calculated since 19 May 2026)

Month	HTML	PDF	XML	Total
May 2026	133	34	8	175
Jun 2026	28	7	4	39
Jul 2026	23	17	3	43
Aug 2026	13	4	2	19

Viewed (geographical distribution)

Total article views: 256 (including HTML, PDF, and XML) Thereof 256 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 18 Aug 2026

Short summary

Measuring airborne particles over water is vital for understanding climate, yet direct observations remain scarce. To address this, we tested a long-distance optical instrument, a large-aperture scintillometer, to measure particle exchange over an inland lake and a coastal bay. Our results demonstrate that this optical method successfully captures particle movements over broad water areas, though dense fog limits its operation. This proves it is a valuable new tool for marine observations.


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