the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Mar 2025
A REtrieval Method for optical and physical Aerosol Properties in the stratosphere (REMAPv1)
Abstract. The stratospheric aerosol is an important climate forcing agent as it scatters some of the incoming solar radiation back to space, thus cooling the Earth’s surface and the troposphere. At the same time it absorbs some of the upwelling terrestrial radiation, which heats the stratosphere. It also plays an important role in stratospheric ozone chemistry by hosting heterogeneous reactions. Major volcanic eruptions can cause strong perturbations of stratospheric aerosol, changing its radiative and chemical effects by more than an order of magnitude. Many global climate models require prescribed stratospheric aerosol as input to properly simulate both climate effects in the presence and absence of volcanic eruptions. This paper describes REMAP, a retrieval method and code for aerosol properties that has been used in several model intercomparison projects (under the name SAGE-3/4λ). The code fits a single-mode log-normal size distribution for a pure aqueous sulfuric acid aerosol to aerosol extinction coefficients from observational or model data sets. From the retrieved size distribution parameters the code calculates the effective radius, surface area density, as well as extinction coefficients, single-scattering albedos and asymmetry factors of the aerosol within the wavelength bands specified by individual climate models. We validate the REMAP using balloon observations after the Pinatubo and Hunga-Tonga Hunga Ha’apai volcanic eruptions, as well as four decades of LIDAR measurements. Within the constraints of a single-mode log-normal distribution REMAP generates realistic effective radii and surface area densities after volcanic eruptions and generally matches the LIDAR backscatter time series within measurement uncertainty. Deviations in aerosol backscatter up to a factor of 2 arise when (non-volcanic) tropospheric intrusions (e.g. from wildfires) are present and their composition deviates significantly from the background type. We describe the products that have been used in CCMI, CMIP6 and other model intercomparison projects, and provide practical instructions for use of the code.
Competing interests: At least one of the (co-)authors is a member of the editorial board of the Geoscientific Model Development. The authors also have no other competing interests to declare.
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 preprint. The responsibility to include appropriate place names lies with the authors.- Preprint
(19240 KB) - Metadata XML
-
Supplement
(1519 KB) - BibTeX
- EndNote
Status: open (until 09 May 2025)
-
RC1: 'Comment on egusphere-2025-145', Anonymous Referee #1, 03 Apr 2025
reply
General comments:
In this paper a method to retrieve stratospheric aerosol properties from observational or model data, named REtrieval Method for Aerosol Properties (REMAP), is described. It is intended especially for use in climate models and model intercomparison projects since the parameters that are most useful for the calculation of the stratospheric aerosol’s radiative forcing such as surface area density (SAD), single-scattering albedo (SSA), asymmetry factor (AF) and the aerosol extinction coefficient (AEC) at any desired wavelength are provided through it. All of the mentioned parameters can be calculated relatively simply from a single lognormal (SLN) size distribution, the retrieval of which is the basis of this method. The paper equips the potential user of the method with a useful generalized scheme to retrieve the SLN distribution (and derived quantities) based on the quality and number of available AEC channels. For this a kind of lookup table of AECs calculated by Mie theory for all combinations of the three parameters describing the SLN (median radius, geometric standard deviation and number density) within physically reasonable value ranges is compared to the AECs of the observational or model data set in a best-fit approach. The data retrieved with REMAP are validated with a long-time LIDAR data set and with in situ balloon-borne measurements after the Hunga Tonga – Hunga Ha’apai eruption. Supplementary material is provided on data sets previously generated using the main method described in the paper, on different satellite data sets and corresponding data quality that they are based on and on data gap filling.
The paper is structured logically and both the science and its presentation are of high quality. The abstract and title characterize the paper well. All methods presented and used are valid. The main method is explained well, in part through a well crafted color-coded overview scheme that makes it easy to follow. The language is concise overall, but some imprecise scientific nomenclature is used. All information that is needed to reproduce the results and implement the method is provided in the paper. The appropriate literature is cited throughout the document. There are some smaller errors scattered throughout the document and mathematical formulae that have to be fixed. Further context should be added in some places.
The paper makes a valuable contribution to modelling science, since it provides climate modelers with a path towards a best guess of the relevant properties for stratospheric aerosol radiative forcing, which is important for climate modelling.
Therefore, I recommend the paper for publication after minor revisions.
Specific comments:
Line 62: The absorption of terrestrial radiation that is also calculated by Mie theory in the process of calculating the aerosol forcing should be mentioned in this sentence.
Line 87: In addition to the mention of Thomason et al. (2008) and maybe even more fitting here would be a citation of Knepp et al. (2024), who show that there is not enough information within SAGE extinction spectra to accurately retrieve bimodal lognormal distributions.
Line 110: You should also mention here in some way, that in order to retrieve AECs from Limb scatter measurements (i.e. from instruments like OMPS-LP, SCIAMACHY, OSIRIS...) the aerosol particle size distribution has to be assumed beforehand. Since you want to learn about aerosol size from the AECs, this is an important, unknown and mostly unavoidable source of error when applying your method to limb scatter measurements.
Equation (1): The logarithm outside of the parenthesis (in the denominator of the fraction) should not be squared. See equation (38) in Grainger (2023). For reference, your σ is “S” in this document.
Throughout the document: The parameter rm that appears in your equation (1) is the “median radius”. The “mode radius” describes the peak location of the size distribution, but since the SLN distribution is asymmetric (in linear radius space) this is not equal to the median radius. Please change “mode radius” to “median radius” throughout the document. Similarly the name “half-width” for σ is not quite correct here, please use “geometric standard deviation” instead.
Sect. 3.1 (a): In the 3-or-more-wavelengths case just the single solution with the lowest difference score D is taken as the solution, correct? There are probably often other solutions with almost as low difference scores. Are the parameter values of these solutions similar or do they scatter strongly?
Line 201 and equation (2): I think it is not correct to call the parameter ς the “standard deviation” as you also talk about the standard deviation of individual data points in line 291 (which does not make sense). I think what you mean is the measurement uncertainty or uncertainty of the data point. In case I am correct please change it to something similar throughout the document.
Line 213 and Figure 2: Here you write that the lowest values of E in Figure 2 are in the core of the Junge layer. But in Figure 2 the lowest values are at around 25 km altitude, and the “core” of the Junge layer is typically more around 20 km and below, depending on which parameter you look at and depending on latitude. So this sentence should be changed. Also, why is this area of lowest errors at such high altitudes?
Line 215-216: Here, you give a reason for the higher errors above 27 km. Please also mention the reasons for the higher errors in the lowest stratospheric region, which you mention in the same sentence.
Line 221-223: These sentences are worded a bit confusing to me. For the Mie calculation only the refractive index (and size distribution) should be needed. RH and Temperatures are needed to calculate the refractive index, not directly for the Mie calculation. Please restructure the sentences to make this clearer.
Section 3.1 (c): Taking both rm and σ from parametrization and only retrieving the number density from a single wavelength would likely lead to very unreliable data. An error in rm or σ individually should already have a potentially strong effect on the retrieved number density, since larger particles scatter much more strongly than smaller particles in this size regime. E.g. a high-bias in rm should lead to a low bias in the number density. Now both other parameters of the SLN distribution are parametrized in addition to likely errors introduced by the SLN assumption itself, which also usually affects the number density the strongest. These errors would of course also propagate to the other quantities you calculate from the size distribution, especially the SAD.
Therefore, it should be at least shortly discussed that this retrieval product based on only one wavelength is attached to a lot of uncertainty and likely more or less unreliable (although it might still be the best possible guess for those periods).
Figure 7: To my knowledge, during the Pinatubo period GloSSAC is based mostly on SAGE II, which had big data gaps in the lowermost Junge layer due to the opacity of the lower tangent heights, but you have continuous retrieval data down to 15 km here. Which measurements are these retrievals based on in this case?
Figure 9: I am surprised about the good agreement for the Pinatubo period due to a few layers of issues: Firstly, there is the SAGE II data gap in the lowermost Junge layer mentioned in the previous comment, which would likely reduce the amount and quality of observational AEC channels available for REMAP. Secondly, you parametrized σ here, effectively only retrieving two of the three SLN distribution parameters. And last but not least, the Pinatubo eruption led to a clear bimodal particle size distribution (as we know from OPC measurements) and you can only retrieve a SLN distribution, which would also be a big source of error. Possibly, the second mode with larger particles dominated the measurement signal so strongly, that the SLN retrieval closely approximates only this second mode and therefore the main scattering signal with the first mode with smaller particles being negligible in terms of scattered radiation?
This is not really a clear question, but if you have thoughts on this, maybe you can specify them in the paper.
Line 426: You state thate there is no satellite extinction coefficient data below ~ 15 km in the 08-12-2022 REMAP profile and that therefore the dotted orange line stays constant. However, It does not stay constant but does show noise. Where is the information coming from here?
Line 430-432: You state that the main contributor to the differences between in situ and REMAP data in Figure 10 (b) is arguably that zonal monthly means of REMAP are compared to the momentary local measurements of POPS. I am not sure if this should be assumed to be the explanation for the substantial difference in altitude of the signal peaks. For this to be the case the plume of Hunga Tonga would have to have been at much lower altitudes at other locations or times within the depicted months, which I don’t know to be the case, especially since this difference is so similar at these two different points in time. The much higher size resolution of POPS as opposed to the SLN retrieval with REMAP could just as well be an important if not dominating factor here. A second mode in the true aerosol particle size distribution would (following von Savigny and Hoffman (2020)) likely lead to an underestimation of the REMAP number density as the AEC spectrum would be more strongly affected by the second mode with larger particles, but it would be captured well by POPS, which could possibly explain the differences in SAD seen in Figure 10 (b).
Line 458-459: Additionally to biases of individual wavelengths and contaminated data points the quality of the product can also strongly depend on the measurement geometry, e.g. solar occultation vs limb scatter measurements, as for the latter AECs can only be retrieved after assuming the particle size distribution.
Throughout the document: In the whole document the name “SAGE III” is used when “SAGE III/ISS” is meant. However, there also was another SAGE III instrument: SAGE III/M3M, that operated between 2002 and 2005 on the Meteor-3M satellite. Please change “SAGE III” to “SAGE III/ISS” in all places, where you don’t write generally about the SAGE III instrument type, but specifically about the SAGE III/ISS instrument.
Also, as a suggestion for future use of the algorithm: SAGE III/M3M should be a very beneficial data set for the use of REMAP both for the retrieval itself and for parametrization, since it covers the same number of aerosol channels as SAGE III/ISS and with overall similar quality. It covers different latitudes however, which may be both an advantage and a disadvantage.
In the supplementary, in S4: Here you argue that the monomodal retrieval can reproduce the aerosol extinction coefficient (AEC) spectrum well that were first calculated from the Wyoming OPC measurements when looking at bimodal PSD data. For the AEC that may be true, however there will very likely be a much stronger bias in the three SLN PSD parameters (median radius, sigma and number density) individually (see for example von Savigny and Hoffmann (2020), which you also cite). There would likely be an especially strong effect on the number density that is stronger the larger the second mode of the bimodal “true” distribution is. This in turn would introduce possibly large errors into other quantities that you calculate, like the surface area density (SAD). Please discuss the possible effects of a wrong SLN assumption in the main text, as it is central to the retrieval method.
Technical corrections:
Line 12: The hyphen in “Hunga Tonga - Hunga Ha’apai” is in the wrong place.
Line 37: “[…] individually by each […]”
Line 44: “[…] (Ambae, … […]”
Line 109: “[…] limb scatter measurements.”
Line 195: “To achieve this, a program […]”
Equation (2): For the two summation symbols, the lambda should sit below them not above or to the side, if you want to write it this way. Better would be a notation without lambdas on the summation symbol, but “i=1” below and “n” above it and then give each lambda outside of the summation symbols an “i” as an index.
Also, please specify in the equation of the weight which β you use here, both should probably also have “obs” as an index.
Line 285: “[…] AEC < 10⁻⁴ […]”, i.e. use superscript here.
Line 286: This sentence makes it sound as if only one SAGE III instrument existed. Please rewrite it in order to distinguish it from SAGE III/M3M, e.g.: “This is exemplified in SAGE III/ISS, which is mounted on the International Space Station, […]”
Figure 6: The title of the x-axis says “wavelength [micron]”, but the data are shown in nanometers.
Line 333: I think you meant to reference those equations here: “[…] from Eqs. (5-7).”, correct?
Line 389: I think this should be “[…] using Eqs. (5-8).”, correct? Since you mention the calculation of the AEC and the SSA in this sentence.
Line 401-403: Maybe add: “We can trace this overestimation back to the lowest altitude levels in the stratosphere, […] ” in order to avoid possible confusion.
Line 405: “[…] even for these poorly […]”
Figure 9 caption: The Raikoke eruption happened in 2019, not 2020.
Line 407: “[…] using an optical particle counter, […] ”
Line 435: “[…] input […] “
Figure 10 (a) – Please mention the altitude range over which the data are averaged (14-30 km according to the text) also in the caption here.
Figure 10 (b): Maybe change the date format so that the month is given as a word, since it is not obvious whether the first number is the month or day here.
Figure 10 caption: missing hyphen in “Hunga Tonga – Hunga Ha’apai”
Figs. 3, 4, 5, 7 and 8: These figures use color schemes that may be very difficult to read to people with color vision deficiencies. Please consider revising these plots with color schemes that are better suited for this, e.g. color schemes with linearly increasing or decreasing brightness such as viridis.
References mentioned:
Grainger, R. G.: Some Useful Formulae for Aerosol Size Distributions and Optical Properties, 2023, accessed August 12th, 2024 at: http://eodg.atm.ox.ac.uk/user/grainger/research/aerosols.pdf.
Knepp, T. N., Kovilakam, M., Thomason, L., and Miller, S. J.: Characterization of stratospheric particle size distribution uncertainties using SAGE II and SAGE III/ISS extinction spectra. Atmos. Meas. Tech., 17, 2025-2054, https://doi.org/10.5194/amt-17-2025-2024, 2024.
Thomason, L. W., Burton, S. P., Luo, B.-P., and Peter, T.: SAGE II measurements of stratospheric aerosol properties at non-volcanic levels, Atmospheric Chemistry and Physics, 8, 983–995, https://doi.org/10.5194/acp-8-983-2008, 2008.
von Savigny, C. and Hoffman, C.: Issues related to the retrieval of stratospheric-aerosol particle size information based on optical measurements, Atm. Meas. Tech., 13(4), 1909-1920, doi:10.5194/amt-13-1909-2020, 2020.
Citation: https://doi.org/10.5194/egusphere-2025-145-RC1
Data sets
REMAPv1 Andrin Jörimann https://www.research-collection.ethz.ch/handle/20.500.11850/713396
SAGE-4λv2 Beiping Luo https://doi.org/10.3929/ethz-b-000714581
SAGE-3λv4 Beiping Luo https://doi.org/10.3929/ethz-b-000715155
REMAP-CCMI-2022-ref Beiping Luo https://doi.org/10.3929/ethz-b-000715176
REMAP-CCMI-2022-sai Andrin Jörimann https://doi.org/10.3929/ethz-b-000714654
REMAP-GloSSAC-2023 Andrin Jörimann https://doi.org/10.3929/ethz-b-000713396
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 77 | 27 | 6 | 110 | 14 | 7 | 6 |
- HTML: 77
- PDF: 27
- XML: 6
- Total: 110
- Supplement: 14
- BibTeX: 7
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 35 | 26 |
| Germany | 2 | 15 | 11 |
| Switzerland | 3 | 12 | 9 |
| China | 4 | 8 | 6 |
| France | 5 | 7 | 5 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 35