the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Performance assessment of the IASI-O3 KOPRA product for observing midlatitude tropospheric ozone evolution for 15 years: validation with ozone sondes and consistency of the three IASI instruments
Abstract. The Infrared Atmospheric Sounding Interferometer (IASI) has been monitoring the atmosphere for operational meteorology and atmospheric composition studies since 2007 with a succession of three instruments aboard the Metop-A (2006–2021), Metop-B (2012-) and Metop-C (2018-) missions. One of the key species monitored is ozone (O3). This study assesses the quality of the regional IASI-O3 KOPRA product, version v3.0, and the consistency of the three IASI instruments, IASI-A, IASI-B, and IASI-C for timeseries and trend analyses. The IASI-O3 KOPRA products for IASI-A, IASI-B and IASI-C. IASI-B show a very good agreement and consistency, better than 1 %, for the tropospheric ozone column (TrOC) and several partial columns (surface-450 hPa, surface-300 hPa) over the three domains, Europe, North America, and East Asia of this study. For the quality assessment and trend analyses, we combine the ozone products derived from IASI-A (2008–2018) and IASI-B (2019–2022) without any bias correction. The comparison with homogenized ozone sondes for six northern midlatitude stations reveals a small negative bias of about 3–6 % of the IASI-O3 KOPRA products in the troposphere for both profiles and columns with rather good correlation between 0.7 and 0.9 and an error estimate about 15–17 % (compared to sondes smoothed with averaging kernels (AKs)). The ozone variability is also well reproduced for all the partial columns with a slight underestimation of about 10 % for the TrOC. Based on the comparison with the ozone sondes, we identified a temporal drift (of about -0.06 ± 0.02 DU/yr in average), while more pronounced in summer, for three different ozone columns (TrOC, surface-450 hPa, surface-300 hPa). However, a significant variability of the estimated drifts depending on the sample of ozone sonde sites is remarked, that does not allow its use for correcting the IASI ozone product timeseries over broad domains. Whereas the upper tropospheric ozone trends are mainly positive or undefined, the lower tropospheric ozone trends are mainly systematically negative. The regions the most affected by negative trends are the Mediterranean, Western North America, Eastern North America and East Asia. Compensations between lower and upper tropospheric trends prevent the identification of any specific long-term behaviour for TrOCs over the three domains. The negative tropospheric ozone column anomalies in the 2020–2022 (post-COVID19) time period observed in our northern hemisphere mid-latitude domains slightly impact the estimated trends but do not change the conclusions stressed before.
- Preprint
(4707 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 26 Feb 2025)
-
RC1: 'Comment on egusphere-2024-4096', Anonymous Referee #1, 31 Jan 2025
reply
General comments:
This manuscript by Dufour et al. provides an insightful performance assessment of the regional KOPRA ozone profile v3.0 product that deserves publication in AMT. The current presentation, however, could be improved regarding scientific clarity and significance. Especially the ‘improvements’ in this version 3.0 with respect to the previous, the retrieval information as captured by the averaging kernels, and the discrepancies in the drift studies are unclear.
Specific comments:
Section 2.1 (1): Briefly explain the main differences of this v3.0 of the KOPRA product with respect to previous versions, especially as the end of Section 3.3.2 mentions that its performances are ‘similar’.
Section 2.1 (2): Why are only morning overpass pixels considered? Please explain.
Line 138: The observation that “the differences can vary at the scale of month or year” does not seem to be a sufficient condition for “preventing assessing a systematic bias between the instruments.” Please explain and rephrase.
Lines 180-183: If the regular smoothing equation is used, just cutting the averaging kernel matrix above the sonde vertical range is equivalent to first extending the sonde profile with the a-priori profile and smoothing over the full vertical range afterwards.
Lines 184-185: “to avoid possible interpolation issues and uncertainties” is quite vague. Is this to conserve the vertically integrated ozone profile or retrieval sensitivity, more specifically?
Line 201: I tend to disagree with the statement that “Outside of this [9-15 km] range, the bias is rather constant in altitude” also given that variations are discussed below. Possibly rephrase?
Line 202-203: “When the sonde profile is smoothed by the AKs, the smoothing errors are removed” This is a brief and rather vague statement for an operation with quite some impact. Better elaborate on this in particular, and on the retrieval sensitivity in general.
Line 206: An obvious validation research question is whether the observed uncertainties match the (prognostic) errors provided within the KOPRA product files?
Figures 4 and 5 should have axis legends and units. Moreover, red and blue colors are mostly saturated in Figure 5, so an extension of the color scale range might be appropriate.
Lines 267-268: “One can also notice that the variability of the TrOC is slightly underestimated by IASI (by about 10%)” Explain how this can be seen from Figure 6. Also explain all aspects of the Taylor diagram in the latter. It is not clear which “curved lines” are referred to in the caption (light or dark grey).
Line 364: “The negative trends reported in the lower troposphere are then not affected by the a priori.” This only holds if the retrieval sensitivities do not significantly change either. Can this be ruled out?
Lines 365-369: Possibly quantify the contribution of drift to the trends?
Line 390: The aim to “provide recommendations for its use in trend analysis” does not appear to me in the text. Does this refer to looking at several tropospheric sub-columns?
The end of Section 3.3.3 rightfully questions the representativeness of the six ozone sonde station data for the three regional studies in this work, especially regarding long-term drifts. The conclusion in line 407, referring to the fact that the mean drift (properly assessed) should indeed not be “largely dependent on the sample of ozone sonde sites”, should be taken at face value and result in a (future) reconsideration of the drift study?
It seems more appropriate to keep the brief yet important Appendices A and B with the main text?
Technical corrections:
Lines 18-19: repetition of IASI-B
Line 51: “remain”
Line 59: “used to study trends”
Line 79: “discusses”
Line 85: Explain “L1C” or leave out.
Lines 89 and 92: Specify which a priori information is discussed, differentiating between temperature and ozone profiles.
Line 116: Provide latitude-longitude definitions for the regions under study.
Line 219: “to calculate”
Line 244: Either “using” or “over” but not both.
Line 255: “errors are removed”
Line 304: “spectral fit”
Table 2: The last row of data (on 300hPa-tropopause columns) seems to be missing.
Figure 8: Match the vertical scales of both plots for better comparability?
Line 434: “derived trends”
Table 3 should be renamed D1 (or similar according to which appendices are kept).
Citation: https://doi.org/10.5194/egusphere-2024-4096-RC1 -
RC2: 'Comment on egusphere-2024-4096', Alistair Bell, 06 Feb 2025
reply
General Comments
The paper shows a validation of retrievals using the KOPRA algorithm from the three iterations of the IASI instrument, which show a low bias between the partial ozone columns for different altitude layers of the troposphere.
The paper also provides analysis of the IASI KOPRA product compared to in-situ sonde measurements. The results show that when sonde measurements are convolved with the KOPRA averaging kernel, the bias is close to zero, and there is an RMSE of around 20% when compared to the KOPRA retrievals. Drifts in retrieved measurements are also compared to ozonesonde measurments, showing a small negative drift, with a low p-value for the tropospheric ozone column.
Ozone trends are then analysed for three regions in the northern hemisphere: North America, Europe and East Asia. It is interesting to see that for the lower troposphere, trends are mainly negative for all regions, with mainly low p-values, whereas for the upper troposphere there is little clear signal except in the East China Sea/Pacific region, where trends are positive. It is also interesting to note that the negative trend in the lower tropsophere remains even when the covid period is excluded from the dataset.
The article presents important results relevant for tropospheric ozone analysis, and should be published in AMT with corrections and clarifications noted below.
Specific Comments
It would be useful to specify more precisely any instrumental changes between IASI A, B and C. If there are no instrumental changes at all, this would also be worth highlighting. In case the performance of the instruments is slightly different, it may also be worth plotting the differences in measuremnent response of retrievals from the three instruments.
It is intriguing that both IASI-A and -C have roughly the same bias compared to IASI-B for all partial columns. Could you specify the overpass time separation of the satellites? As the rate of ozone formation and depletion could be quite high at the local time for which observations are made, I wonder if the time separation of the satellite could explain this bias?
line 175: “The coincidence criteria used for the validation are ±1° in latitude and ±1° in longitude around the sonde station, a time175 difference shorter than ± 6 hours” - can you confirm that there was no systematic change in the time difference between iasi/sonde observations?
Section 3.3.1 It would be useful to compare these results with other IASI ozone retrievals, such as those from Boynard et al. (2016, 2018), to note whether the biases and RMSE values observed in here are consistent with other retrieval algorithms
line 208: “The largest differences and RMSE in the first two kilometers when comparing to raw sondes are likely due to an issue because high altitude stations (Payerne, OHP, and Boulder) are mixed with low altitude stations.” - It is not clear to me why this would create a large bias and rmse, more likely your next point about the lack of sensitivity of satellite observations near the surface.
figure 5: This is a nice plot, but seems to be aiming to show different things that could perhaps be better expressed with other plots. For example, to show the drift of IASI retrievals, the raw/smoothed sonde retrievals minus IASI retrievals could perhaps be plotted over time. This might more clearly show a drift. Where the negative anomolies are highlighted in 2020 for example, the problem is that from the combination of sonde datasets, it is not clear how many profiles from each location are used at which time. It may change conclusions if anomolies are not uniformly distributed across all locations.
Figure 9: Interesting to see that the trends in the Mediterranean sea, Bay of Biscay and more Southernly part of the North Atlantic Ocean all have negative trends with low P-values, whilst the East China Sea/Pacific have postive trends with low P-values. Are there changes in ozone precursors that could have caused this?
Technical Corrections
line 17: Please define KOPRA before first usage
line 20: please correct to make clear: “across the three study domains: Europe, North America, and East Asia.” (or similar)
line 38: ...in addition to being...
line 55: were more likely negative
line 59: ...for trend studies...
line 95: lower than?
Figure 3: I found it quite hard to find and guage the size of some circles on this plot - I wonder if it meets requirements for visual impairments. Perhaps it would be better to remove the topographic shading?
Figure 4: Plots require axis labels and units
Figure 5: again no y axis label
Figure 6 (right): I’m not sure if I see the Boulder star. Maybe you can add more colours to the plot to distinguish stations?
Figure 8: is TOC the same as TrOC?
line 419: “In this context of uncertain trends and opposite behavior in the lower and upper troposphere which likely compensate for the TrOC, the questions about possible drifts, more pronounced in summertime, between our sample of ozone sonde time series and the IASI retrievals should be investigated in more detail.” - this sentence should be rephrased or broken down into two sentences.
Citation: https://doi.org/10.5194/egusphere-2024-4096-RC2
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
74 | 14 | 5 | 93 | 1 | 1 |
- HTML: 74
- PDF: 14
- XML: 5
- Total: 93
- BibTeX: 1
- EndNote: 1
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1