the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Feb 2025
Extension of the S5P-TROPOMI CCD tropospheric ozone retrieval to mid-latitudes
Abstract. Tropospheric ozone, a key atmospheric pollutant and greenhouse gas, exhibits significant spatio-temporal variability on seasonal, inter-annual, and decadal scales, posing a challenge for satellite observation systems. Methods like the Convective Cloud Differential (CCD) and Cloud Slicing Algorithms (CSA) are standard for Tropospheric Column Ozone (TCO) retrievals but are limited to the tropical band (20° S–20° N). This study presents the first successful global application of CCD retrieval outside the tropical region. We introduce the CHORA-CCD (Cloud Height Ozone Reference Algorithm-CCD) for retrieving near-global TCO from TROPOMI. It utilises a local cloud reference sector (CLCD, CHORA Local Cloud Decision) to determine the stratospheric (above cloud) column ozone (ACCO). The ACCO is subtracted from the total column in clear-sky scenes to determine the TCO. The new approach presented here minimises the impact of stratospheric ozone variability, which is generally higher in the extratropics. An iterative approach is used to automatically select an optimal local cloud reference sector around each retrieval grid box, varying the radius from 60 to a maximum of 600 km, for which a mean TCO is determined until a sufficient number of ground pixels with nearly full cloud cover are found. Due to the prevalence of low-level clouds in mid-latitudes, the TCO calculation is constrained to the column from the surface up to the reference altitude of 450 hPa. There are two independent methods used: (I) CLCD-C, which uses an ozone climatology and (II) CLCD-T, an alternative method which estimates the ACCO at 450 hPa by linear regression (Theil-Sen) in cases where the cloud-top-heights in the local cloud sector vary sufficiently. The Theil-Sen approach is a combination of the CCD and CSA methods. The CLCD algorithm dynamically decides between the CLCD-C and CLCD-T to determine ACCO depending on the cloud characteristics. The CLCD algorithm is further refined by introducing a homogeneity criterion for total ozone to overcome inhomogeneities in stratospheric ozone. Monthly averaged CLCD TCOs have been determined over the tropics and mid-latitudes (60° S–60° N) using TROPOMI data from 2018 to 2022. The method’s accuracy was investigated by comparing spatially collocated SHADOZ/WOUDC/NDACC HEGIFTOM ozonesonde measurements from 36 stations. The validation results reveal that CLCD TCO retrievals are in good agreement with ozonesondes at most stations with an overall statistical bias of 0.6 DU and dispersion of 2.5 DU. Across all stations, the maximum bias and dispersion are around ∼5 DU and 4 DU, respectively. The CLCD approach effectively captures tropospheric ozone enhancements across diverse regions, including Northeast China and North America, with particular sensitivity to areas impacted by significant emission sources. Our results demonstrate the advantage of using the modified local cloud reference sector, providing an important basis for subsequent systematic applications in current and future missions, in particular, geostationary satellites with an emphasis on observing higher latitudes.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Measurement Techniques.
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
(17340 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2025-306', Anonymous Referee #1, 03 Apr 2025
reply
Manuscript Review - Maratt Satheesan et. al, 2025
Journal: EGUsphere
Date Submitted: 2 April 2025
Title: Extension of the S5P-TROPOMI CCD tropospheric ozone retrieval to mid-latitudes
Authors: Maratt Satheesan, Eichmann, Weber, Van Malderen, Stauffer, and Tarasick
Summary:
This paper presents the first application of a global convective cloud differential (CCD) approach outside of the tropics, where the standard Cloud Slicing and CCD algorithms have been used for decades with limited geographic range (i.e., confined to the tropics). The multi-year mean product produced with the new algorithms agree well with the established approaches in the tropics. The advantage of the new approach lies in its ability to produce results outside of the tropics, where clouds tend to be lower. Having such a product validated and regularly available would be of benefit to tropospheric ozone trend analyses and the identification of influences from large scale biomass burning events.
In its current form, I found this paper challenging to read and review. Overall, the paper could stand some tightening up and a sharper focus on the results that the authors wish to communicate. At present, I found the paper too long, often repetitive, and too broad in scope.
Up front in the paper, it would be nice to know:
- What product does this new approach produce? With what temporal and spatial resolution?
- Under what conditions is this new approach useful? Seasons? Cloudy/clear scenes? Latitudes?
- Over what time period must results be averaged to be useful/reliable?
- What are the uncertainties in the product and how are they impacted by geographic locale, temporal changes, and meteorological conditions?
In its present form, I feel unclear on what the authors are trying to communicate in this paper. Is this a techniques paper or is there science to be gleaned from their analysis?
I had trouble following the use of cloud free and cloudy regions to determine a TCO. And I’m not sure how the authors’ definition of the TCO – surface to 450 hPa – relates to prior definitions for comparison of the product. What fraction of tropospheric ozone is found between the surface and 450 hPa vs the column from the surface to the tropopause?
In addition, in the Section 5 - Validation and discussions - far too often, the authors make suppositions and call out “possible” and “likely” explanations for characteristics in their data without or with meager evidence. Their claims seem stronger than their proof (if it is presented). I would like to see more connecting of the dots to demonstrate their claims match the evidence.
I also have concerns about the error budget formulation and how this algorithm functions in regions of topographic variability. Please look at my detailed comments below.
Overall Recommendation:
I recommend against publishing this manuscript in the current form. A major rework that addresses my concerns (and those of the other reviewers) would improve the draft – if resubmitted, I would be willing to re-review a revised manuscript.
Detailed Comments (by line numbers):
Abstract: Consider shortening and focusing the abstract on the key results and takeaways for the reader. Remove background information.
31: “It empahsises…” To what does “it” refer?
64: reference is made to a “homogeneity criterion,” which I know is defined later in the paper. At this point, I am wondering what that means, however.
68-69: “e.g., long-range transport of boundary layer air to the upper troposphere.” I’m not sure I completely understand this example. I think of air moving to the UT/LS region vertically quite rapidly in conditions of convective instability. In the UT/LS, it can be transported long distances. Perhaps the authors could rephrase to clarify the meaning.
87: “August 2019 (Ingmann et al., 2012).” Curious about the two dates here, with the reference to a paper 7 years prior to the date of application?
117-118: regarding the conversion of ozone mixing ratio provides to columns in DU, the authors reference their own 2024 paper. Is the authors’ approach different than those used previously? Is it novel? Does it agree well with prior approaches? I think there’s probably a primary reference to cite here for this process…
118: You mention spatial collocation conditions for ozonesondes and TROPOMI data. What about the temporal coincidence criteria?
132: You reference McPeters et al, 2003, which appears to be an AGU presentation. I think the paper you want is McPeters et al. (2007). Also, is there something newer than this? The data that went into the McPeters et al. climatology is 20+ years old. Do you have any concerns about trends/drifts over the intervening years?
138: I think you should probably have a reference to Figure 2 somewhere around this point.
140-141: You require a cloud-top-height greater than 2 km. Is this above sea level or about ground level?
141-142: You justify a circular reference area for the extratropics due to varying wind directions and mention that there is “no single dominant direction.” Do you have a reference to support this assertion or data to show? Have you tried different shaped regions? Might the shapes depend on the meteorological conditions? I can imagine a frontal boundary crossing through one of these regions with very different tropospheric ozone ahead of and behind the front. Have you tested the algorithms performance in such cases?
146-147: “...the common reference level or upper altitude limit for the tropical ozone column is 450 hPa (~5.5 km) instead of 270 hPa.” I’m a bit confused by this statement. I thought the clouds were higher in the tropics, so I am surprised the statement references a lower altitude (higher pressure) for tropical ozone column. Is this a typo?
163: regarding the daily averaging in each grid box: the Sentinal-5P overpass is at 1330 local time roughly. I guess some 0.5X0.5 grid boxes could have successive orbits in them...With a scan width of 2600 km (2.5 deg of latitude or >2.5 deg of longitude) and a pixel at 7 km X 3.5 km2 in the middle, does the introduction of successive scans negatively impact the value assigned to a grid cell? Would you be better off using only pixels from the same overpass/scan, then interpolating in space between successive scans rather than computing a daily average?
165-166: you assign a bad quality flag to negative data and substitute a “designated fill value” in the data set. Is information lost by not preserving the negative values? The distribution should be informative as to the performance and reliability of the approach. How did you handle the error calculations for the data product in regions where there were negative values? Should not these negative values be included in the statistical analysis, including means? By eliminating the negative values, are you artificially biasing your data set to the high side?
168 - 170: “TCO is computed only when…otherwise it is replaced with a fill value and assigned a quality flag of 5.” How was this criterion determined? I'm not sure how to reconcile its application with the procedure outlined in the prior paragraph which requires clear sky conditions? Also, again, how does this treatment of the data impact your uncertainty estimates?
Figure 2: the Boulder ozonesonde station – This scene has significant variations in surface altitude. This box has altitude variations of ~100 m in the "flat" part to the east and 0.5 km to the immediate west with differences up to 3 km vertically just a bit further west but within 30 km! How does this algorithm work with such changes in topography?
175ff: I am unclear at the formulation of this error budget. I assume these are calculated on the same 0.5X0.5 grid? Additionally, if this is a total uncertainty in the daily tropospheric column, but is derived from measurements made essentially only at 1330 local solar time, how do you account for variation within the troposphere throughout the rest of the day?
211: what is the temporal coincidence criterion for matches with ozonesondes? All the maps in the paper show multi-year averages. Can you compute a monthly average with a subset of the CLCD TCO that matches the days on which the ozonesondes were flown and the CLCD worked?
221ff: You show the results for Lindenberg, Uccle, and Payerne. These stations are within 800 km of each other. Thus, I do not have confidence in your statement that “these results imply that the CLCD method performs consistently across midlatitudes.” This sample is too small and too geographically close together.
225: Again, “these values highlight the overall reliability of the method.” I do not believe you have provided enough evidence to support this assertion. I am also unclear as to what CLCD product is being validated: daily maps, monthly means, multi-year means?
235-236: “Additionally, CLCD effectively captures seasonal variations in tropospheric ozone across most stations.” Define effectively? The magnitudes of the seasonal cycles of CLCD results shown in Figure 7 look different from the actual data, although given the relatively large uncertainty in the CLCD product, not statistically significantly so…
238: “...more reliable ozone measurements across diverse atmospheric conditions.” Can you show results for how the dynamic choice between CLCD-C and CLCD-T improved the outcome? What would have happened if you would have stuck with one or the other?
243: “We conclude no dependencies” in Figure 8. Not sure. The purple dots show some correlation for TOZ, ACCO, and cloud top height, to my eye. Did you run statistical tests on these data or subgroups of data?
268: “The higher cloud cover…” You mean “greater amount of cloud cover” rather than “higher altitude cloud cover,” correct?
Figure 7: Seems as though at most sites, the annual cycle has a greater magnitude in the CLCD data than in the sonde data. Does that provide insight into the approach itself? Also, the caption describes “ozonesonde tropospheric columns above 450 hPa.” Is it below in altitude and above in pressure coordinates?
Figure 8: Same comment about directional reference in the figure caption. The caption indicates “ACCO up to 450 hPa.” Does this mean up to in pressure coordinates or up to in altitude coordinates?
272-273: I gather from these statements that you are applying a climatological correction when the cloud tops are not at 450 hPa. Is that correct? How does this impact comparisons with other TCO products?
281-282: “...the high positive bias in TCO is most likely caused by low-level clouds.” Is this a general statement or something specific to Yarmouth?
291ff: I am having trouble following the arguments in this paragraph.
311: Regarding the ITCZ, the authors state agreement is “likely due to the presence of higher cloud-top-heights.” Can the authors demonstrate this assertion is true?
317-318: “Costa Rica [shows] both the highest positive bias and the largest correction…” What did co-author Ryan Stauffer have to say about this data set? I believe he is the NASA PI for the station.
354-355: In explaining Hanoi, you identify “transport process…likely contributes” and “It is probable that…” Please show some evidence that these are the factors that influenced your results.
360: Again, this time for Ascension, “likely tied to biomass burning…” without proof. How can you demonstrate this impact.
364: Again for Ascension, “These clouds likely contribute…” Ditto.
375: “This might explain the high spread during spring in almost all the high latitude NH stations…compared to the SH station.” I am having difficulty looking at the figure and discerning to what the authors are referencing. Also, the NH vortex in winter/spring is often displaced from the pole, causing significant zonal gradients that do not appear as often in the SH where the vortex during winter/spring is usually centered on the pole.
380: “The persistent scatter is consistent across different seasons, indicating substantial variability in stratospheric ozone.” I am not certain that follows logically, even if it is true. Again, please demonstrate.
384-385: For King’s Park, “This suggests substantial ozone transport…resulting from stratospheric instructions and biomass burning…” Could be. Please show some evidence.
517: “Despite low NO2 concentrations, elevated tropospheric ozone levels…” The boundary layer ozone chemistry is complex and non-linear: moving from a NOx limited to a hydrocarbon limited environment changes the production/loss balance. Furthermore, while NO2 may well be mainly confined to the boundary layer in these satellite retrievals, O3 is probably not. I would hesitate to assign high TCO with high surface concentrations, or assign low NO2 with high TCO as indicating long-range, free-tropospheric transport without surface measurements. I realize you are citing other studies that have looked at this more carefully, but I am not sure why this is in your paper. I find myself unconvinced by any of the “Selected applications” of section 6.
Citation: https://doi.org/10.5194/egusphere-2025-306-RC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 193 | 41 | 11 | 245 | 22 | 25 |
- HTML: 193
- PDF: 41
- XML: 11
- Total: 245
- BibTeX: 22
- EndNote: 25
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
