the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Validation of 12 years (2008–2019) of IASI-CO with IAGOS aircraft observations
Abstract. IASI-A, B and C (Infrared Atmospheric Sounding Interferometer) are nadir looking thermal infrared sensors which are monitoring the atmospheric composition since 2008. Atmospheric Carbon monoxide (CO) is retrieved from IASI radiances with two algorithms: the SOftware for a Fast Retrieval of IASI Data (SOFRID) and the Fast Optimal Retrievals on Layers for IASI (FORLI). The airborne in-situ observations from the In-service Aircraft for a Global Observing System (IAGOS) European Research Infrastructure have been used to validate the IASI CO retrievals. The validation study of IASI CO data performed in 2011 whith IAGOS data was limited to two airports (Frankfurt and Windhoek) and 2 years because of the limited sampling at the other IAGOS sites. The extension of the IAGOS infrastructure during the last decade enables a validation with enough temporal sampling at 33 airports worldwide over the whole IASI-A period (2008–2020).
The retrievals provide between 1.5 and 3 independent pieces of information about the CO vertical profile and we have selected to validate the surface-600 hPa and 600–200 hPa partial columns in addition to the total column. The ability of the retrievals to capture the CO variabilities is slightly different for the two retrieval algorithms. The correlation coefficients are generally larger for SOFRID, especially for the total and lower tropospheric columns, meaning a better representation of the phase of the variability, while the amplitude of the variations of FORLI are in better agreement with IAGOS in the mid-upper troposphere. On average SOFRID and FORLI retrievals are underestimating the IAGOS total columns of CO (TCC) by 8±16 % and 6±14 % respectively. This global TCC agreement between the algorithms is hiding significant vertical and geographical differences. In the lower troposphere (Surface-600 hPa) the bias is larger for FORLI (-11±27 %) than for SOFRID (-4±24 %). In the mid-upper troposphere the situation is reversed with a bias of -6±15 % for FORLI and of -11±13 % for SOFRID. The largest differences between the retrievals are detected south of Bangkok where SOFRID underestimation is systematically larger for the TCC and mid-upper tropospheric column. North of Philadelphia FORLI biases are significantly larger than SOFRID ones for the TCC and the lower tropospheric columns. Our validation results will provide a better characterisation of IASI-CO data to the users and help improve the retrievals for future versions.
- Preprint
(22486 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 24 May 2024)
-
RC1: 'Comment on egusphere-2024-30', Anonymous Referee #1, 15 Apr 2024
reply
The IAGOS programme is unique in providing routine observations of the vertical profiles of CO, O3 and other compounds in the troposphere and UT/LS. As such it is a key source of information to evaluate profiles retrieved from satellite sensors. The paper of Barret et al. is, to my opinion, an important addition to the existing literature on IASI-CO retrievals and validation efforts. It is well written and was a pleasure to read, providing insights in the two retrievals and how they compare with IAGOS. As such I am in favour of publishing these results, after my comments below have been addressed.
General comment:
The other important source of validation data are the surface remote sensing FTIR instruments (NDACC). These are mentioned briefly but I would like to read more about the comparisons with the findings from these evaluations (e.g. EUMETSAT validation report). Are the main conclusions similar? Is the underestimate reported for the TCC quantitatively in agreement with FTIR evaluations?
Abstract:
l 5: It would be good to mention that MLS CO is used for the stratospheric part.
l 9: Does the period "2008-2020" cover 12 or 13 years?
l 12: What is meant by "to capture the CO variabilities"? Variability in time, or in the profile? I suggest to replace "variabilities" by "variability".
l 13: Same question for "correlation coefficients". Does this refer to temporal correlations or vertical profile correlations?
l 21: The last sentence "Our validation results will provide a better characterisation of IASI-CO data to the users and help improve the retrievals for future versions" can be formulated better. I would replace "will provide" by "do provide". What does "better characterisation of IASI-CO data to the users" mean? Please reformulate "improve the retrievals for future versions".
l 36: Remove the "(" before "Hurtmans".
l 47: The introduction gives the impression that the study of De Wachter 2012 is the only comparison with IAGOS. It would be good to mention that George 2015 also includes comparisons with IAGOS profiles, although this comparison is somewhat limited.
Introduction: The CO retrievals are introduced with a good set of references. But I am missing a paragraph in the introduction on IAGOS. What is it, what can be measured, some key achievements and key references, including validation work, to provide the reader with a background and further reading.
Sec 2.1: The key reference for SOFRID is De Wachter. But this paper is more than 10 years old. Has there been any CO retrieval development in the meantime based on SOFRID? Are the updates made to SOFRID CO documented somewhere in more detail (is there a recent ATBD)?
l 70: "instead of operational EUMETSAT Level 2 IASI products". Why?
l 71: Could you please motivate why "The noise of the measurement covariance matrix has been reduced from 1.4 to 1.0 · 10−8 W/(cm2 sr cm−1)" I assume that the performance of IASI has not improved over time?
l 80: Same question as for SOFRID. The key ref for FORLI is from 2012 as well, again more that 10 years old. I noted the ATBD is from 2014. Are there updates compared to the 2012 Hurtmans paper? Any relevant evaluations of the FORLI CO retrieval published after 2012? Does FORLI v20151001 introduce any important changes compared to Hurtmans 2012?
l 97: What is the reason that "only airports providing at least 60 days with valid data " were selected? Is a number of 60 linked to the quality of the comparison? Even just one profile can still provide a useful comparison.
l 122: The difference in DFS is striking. The statement that "the reduction of the noise of the measurement covariance matrix relative to De Wachter et al. (2012)" is partly responsible askes for some more explanation. How can the noise be reduced by such a large factor? How does this compare to the a-priori noise assumed in FORLI? Please provide more detail.
l 119, 133: I note that TCC is used in two ways, either as total column (retrieval of 1 quantity) or as total atmosphere (for DFS = 2.9). Maybe better to use "Total atmosphere" instead of TCC in table 1.
l 177: I find the statement "because they provide the best assessment of the real differences between the in-situ and the remote sensed data" a bit dubious. Equation 1 describes how the retrieval relates to the real profile, and provides the best way of comparing. This is also evidenced by the validation results, e.g. l 238.
l 297: This mentions "two major updates of EUMETSAT Level 2 data processing". Could you please provide details on these updates in Sec. 2.
l 377: "timeseries"
l 387: What does "not statistically significant" refer to? From the paper I got the message that TCC globally is not really different between SOFRID/FORLI, but that negative biases are observed.
Citation: https://doi.org/10.5194/egusphere-2024-30-RC1
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
136 | 39 | 11 | 186 | 5 | 5 |
- HTML: 136
- PDF: 39
- XML: 11
- Total: 186
- BibTeX: 5
- EndNote: 5
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1