the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Jan 2025
Improvements in aerosol layer height retrievals from TROPOMI oxygen A-band measurements by surface albedo fitting in optimal estimation
Abstract. The Aerosol Layer Height (ALH), from the Sentinel-5P/TROPOMI L2__AER_LH product, is based on an optimal estimation (OE) approach, fitting cloud-free measurements to synthetic reflectances in the strongest oxygen absorption band, provided by a neural network trained with high resolution simulated reflectances. The ALH has been continuously improved since its release in 2019, focusing especially on (bright) land surfaces, over which the ALH product showed underestimated aerosol layer heights (biased towards the surface). This paper describes the latest updates of the ALH product, that includes first the introduction of the Directional Lambertian-Equivalent Reflectance (DLER) climatology to improve the surface albedo characterisation over land. Second, the paper describes a further improvement, adding the surface albedo in the feature vector of the OE inversion, using the DLER as prior information. Using this approach, the retrievals over land largely match the retrievals over ocean, which have shown a good comparison with validation data since its release, most notably with CALIOP weighted extinction heights. The albedo is fitted for both land and ocean surfaces, but the implementation is different over land and ocean because of the large range of land surface albedos. Over land, the a priori surface albedo values are relaxed so the fitting procedure can incorporate the albedo effects in the retrieval over land. Over ocean, the retrievals are optimised by tuning the a priori error settings. The current implementation improves retrievals over land with about 1.5 times more converged results, and decreases land-ocean contrasts in the aerosol layer height retrievals. The average difference between CALIOP weighted extinction height decreased for selected cases from about −1.9 km to −0.9 km over land and from around −0.8 km to +0.1 km over ocean.
- Preprint
(23028 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on amt-2024-198', Anonymous Referee #1, 29 Jan 2025
Review of “amt-2024-198”
Overall this is a significantly well-written article, clear and logical in its premises and in the material presented. The topic described is perfectly adherent to the purpose of AMT as it describes the improvements achieved in deriving the height of aerosol layers by including the surface in the inversion of TROPOMI measurements of the oxygen band.In the spirit of improving its scientific impact, I list specific comments below. In general, these are minor corrections that should not cause much effort. I would appreciate more contrast and integration with the scientific findings of three relevant papers focusing on retrieval of aerosol layer height. I believe that this would strengthen not only the present paper but also help the community of interested developers and users to advisely both use the product and implement next endeavours. The three papers are Sanders et al (2013, 2015) and Kylling et al (2018).
Sanders, A. F. J. and de Haan, J. F.: Retrieval of aerosol parameters from the oxygen A band in the presence of chlorophyll fluorescence, Atmos. Meas. Tech., 6, 2725–2740, https://doi.org/10.5194/amt-6-2725-2013, 2013.
Sanders, A. F. J., de Haan, J. F., Sneep, M., Apituley, A., Stammes, P., Vieitez, M. O., Tilstra, L. G., Tuinder, O. N. E., Koning, C. E., and Veefkind, J. P.: Evaluation of the operational Aerosol Layer Height retrieval algorithm for Sentinel-5 Precursor: application to O2 A band observations from GOME-2A, Atmos. Meas. Tech., 8, 4947–4977, https://doi.org/10.5194/amt-8-4947-2015, 2015
The exact questions are:
- From the contrast with the findings of Sanders et al., can the authors answer whether the surface shall eventually be always fitted?
- From the contrast with the findings of Kylling et al (2018) can the authors make an effort and compare their knowledge of the problem at hand with the discussion points provided in Section 4 of that paper? There, other algorithms based on the fit of the oxygen A band have been compared and some open points have been left unanswered.
Specific comments
P2 L53: here it seems reasonable to add the Kylling et al paper (https://doi.org/10.5194/amt-11-2911-2018).
P3 L79: it will be interesting to understand if and why the conclusions by Sanders et al 2015 are confirmed or not by this paper.
P4 L115: please add reference to the Kylling er at (2018) paper for consistency. There also other algorithms are described using complementary techniques.
P4 Section 2.1: I am not sure that the title of this section is fully appropriate.
The first paragraph presents the most important specifications of the sensor, but the second presents the details of the algorithm. I suggest a more descriptive and appropriate title, if the authors do not want to add a separate section.
About the algorithm and its details: there is an omission in the description and that is the spectroscopy used and especially what form of line (e.g. Voigt, Gaussian, Rautian, speed-dependent and so on) was used to recreate the oxygen band. I invite the authors to provide details. I ask this because depending on the height of the tropospheric column where we place ourselves, one line shape will be more appropriate than others. I am well aware that this is beyond the scope of this paper, but for future reference it is interesting to know how the authors set up the RT calculations.
P5 L132: Can the authors justify the choice of the Henyey-Greenstein function and the value of asymmetry parametr? First, the advantage of a Mie ot T-Matrix over the H-G phase function is that they better describe aerosol particle scattering. Even more important is the interrelation to the size distribution. Size distribution, asymmetry parameter and single scattering albedo determine the backscattering efficiencies of the particles.
For clouds and a g=8.44, for instance, a H-G phase function causes a 60% deviation in backscatter when compared to a Mie phase function (Hansen 1969). For aerosols in accumulation mode, the adoption of H-G gives rise to a 12% discrepancy against Mie-based approach. (Marshall et al., 1995)
Moreover, assuming that for typical TROPOMI line-of-sights we are in the backward scattering direction, the H-G phase function would clearly understimate the signal (Fig. 1 in Seidel et al, 2010). This result is in line with the findings of Marshall et al, because the H-G implies the overestimation of the asymmetry parameter therefore understimating the aerosol signal for the same measurement at TOA.
J. Hansen, “Exact and approximate solutions for multiple scattering by cloudy and hazy planetary atmospheres,” J. Atmos. Sci. 26, 478–487 (1969)
Stephen F. Marshall, David S. Covert, and Robert J. Charlson, "Relationship between asymmetry parameter and hemispheric backscatter ratio: implications for climate forcing by aerosols," Appl. Opt. 34, 6306-6311 (1995)
Seidel, F. C., Kokhanovsky, A. A., and Schaepman, M. E.: Fast and simple model for atmospheric radiative transfer, Atmos. Meas. Tech., 3, 1129–1141, https://doi.org/10.5194/amt-3-1129-2010, 2010.
P6 L161: For the casual reader, please rephrase or shortly explain what “Pre-whitening is applied” mean.
P6 L176: “In section 3.3 the effect of different weights in the a priori error covariance matrix is described for retrievals over ocean.”
Why only over the ocean and not also for pixels over land, since the authors state these are the cases where they see a gain in accuracy (P4 L94)? I find inconsistent to show the weights for ocean pixels, which have lost accuracy in some cases, while not showing those for land pixels, which are the true improvement of this version of the algorithm.
P6,7 L187-189: “Note that the ALH does not take different aerosol types into account, but assumes weakly absorbing aerosols, because in the O 2 A-band the penetration depth is controlled by the scattering of the aerosol layer, not the absorption.”
Fair enough. But again, the scattering part in extinction needs to be correctly assumed. See my comment above about the H-G phase function and size distributions. The variety of aerosol types you analyse cannot be captured by a single aerosol model.
P7 Eq 4: Can the authors justify the choice for this definition of aerosol layer height? (see Section 2.1.1 in Kylling et al, 2018)
P13 L300: In fact, Figure 4 is even more convincing than Figure 3 in demonstrating the algorithm's improvements. The authors should also not overlook the fact that CALIOP itself is not free from errors arising from the assumption of a lidar ratio that, especially in cases of high optical thickness, does not describe properly multiple scattering (see e.g. Cuesta et al., 2009, 2015).
This consideration naturally leads me to manage an expectation of my own, which can be summarized in the following question: How does the algorithm behave as a function of the optical thickness of the aerosol layer?
I would like the authors to develop the issue and answer the question above and also this one: how does the addition of the surface in the state vector correlate with the accuracy of the aerosol layer height via the surface-AOT correlation in the oxygen band continuum (i.e., for wavelengths shorter than 758/9 nm)?
I suspect that what is gained in fitting the surface is lost in determining AOT, which parameter becomes a de facto error sink.
For the avoidance of doubt: I am not asking to validate the AOT derived from the oxygen band. I am clear about its limitations and that a multi-spectral approach is more appropriate. What I am asking is to inspect the trends between TROPOMI AOT_O2, ALH accuracy (TROPOMI - CALIOP) and goodness of the surface fit.
Cuesta, J., Marsham, J. H., Parker, D. J., and Flamant, C.: Dynamical mechanisms controlling the vertical redistribution of dust and the thermodynamic structure of the West Saharan atmospheric boundary layer during summer, Atmos. Sci. Lett., 10, 34–42, https://doi.org/10.1002/asl.207, 2009
Cuesta, J., Eremenko, M., Flamant, C., Dufour, G., Laurent, B., Bergametti, G., Höpfner, M., Orphal, J., and Zhou, D.: Three-dimensional distribution of a major desert dust outbreak over East Asia in March 2008 derived from IASI satellite observations, J. Geophys. Res.-Atmos., 120, 7099–7127, https://doi.org/10.1002/2014JD022406, 2015
Typos and style
P6 L161: “to scale the elements of the state vector elements”. Perhaps a repetition?
P6 L181: “a set of nice different cases”. Perhaps nine?
P6 L186: “CALIOP L1 data”. Spurious sentence.
Citation: https://doi.org/10.5194/amt-2024-198-RC1 -
RC2: 'Comment on amt-2024-198', Anonymous Referee #3, 12 Mar 2025
The authors present a new aerosol layer height (ALH) product from TROPOMI oxygen A-band measurements. This product employs surface albedo estimated from TROPOMI measurements (as opposed to previous versions that used GOME-based albedo data). This is an important change as it enables usage of the correct viewing geometry (TROPOMI makes measurements in the afternoon while GOME does so in the morning) and hence provides the proper directional reflectivity. Comparisons with CALIPSO measurements demonstrate significantly improved ALH retrievals over land (including bright surfaces) and decreased land-ocean contrast. This work is novel and well written and certainly publishable in AMT.
I have only two major comments.
First, it is mentioned that the neural network training was done assuming fixed aerosol properties, in particular a single scattering albedo of 0.95 and a Henyey-Greenstein function with asymmetry parameter 0.7. Does this not bias the retrievals when the actual aerosols present have different properties? The authors note that “the ALH does not take different aerosol types into account, but assumes weakly absorbing aerosols, because in the O2 A-band the penetration depth is controlled by the scattering of the aerosol layer, not the absorption.” I am not so sure that this is true (and even if it is, the asymmetry parameter would matter). The single scattering albedo and the phase function do affect the relative interplay between aerosol scattering and gaseous absorption. At the very least, the authors should do some sensitivity studies (varying SSA and asymmetry parameter) to prove their hypothesis that the absorption does not matter.
Second, some of the aerosol plume events have an hour or longer time difference between the Sentinel-5p and CALIPSO overpasses. What was the purpose of selecting these cases? There needs to be some text describing the rationale for the case selection.
Minor comments/typos:
Line 37: space-based instruments -> space-based retrievals
Line 42: extend -> extent
Line 54: remove “like”
Line 65; Line 104: wavelengths -> wavelength
Line 114: 20204 -> 2024
Line 123: For the ALH -> For the ALH retrieval,
Line 174: maximum likelihood -> maximum likelihood estimate
Line 185: remove “CALIOP L1 data”
Lines 212-213: land surfaces and ocean surfaces -> land and ocean surfaces
Line 303: weakly scatterers -> weak scatterers
Citation: https://doi.org/10.5194/amt-2024-198-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 150 | 33 | 7 | 190 | 7 | 6 |
- HTML: 150
- PDF: 33
- XML: 7
- Total: 190
- BibTeX: 7
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1