General comments

The study conducted by Tang et al. evaluated the impact of assimilating multispectral/joint versus TIR-only retrieval, column versus profile retrieval, and joint products versus single-spectral products separately, through a 15-day inversion in CAM-chem+DART system, based on MOPITT CO database. They evaluated the assimilation performance against both the assimilated data and independent data (TROPOMI, TCCON, NOAA CCGG, IAGOS, and WE-CAN). Overall, this study is convincing and of great interest to the community studying CO or satellite retrieval, within the scope of AMT.

Readers would be benefit from understanding if such comparison remains consistent across other seasons and years. In section 7.4, the authors acknowledge that these results might not be representative for other seasons, and performing a long-term assimilation has high computational cost. Please consider a direct comparison of measurements, rather than their assimilation performance, to evaluate the potential of robustness in discrepancies and similarities across datasets over varying timeframes.

Please clarify the treatment of observation error estimates. Did the inversions include the information of observation uncertainty? Can the observation uncertainty provided by MOPITT associated with measurements, effectively highlight the useful information from each dataset?

Specific comments

Figure 3: Please explain the criteria used to reject observations too far from the ensemble mean. Providing details on the threshold or methodology employed for this rejection would enhance the transparency of the assimilation process.

Figure 5: The authors did a good job in effectively presenting the inversion performance in Figure 5, but did not discuss the 200 hPa results in the main text.

Figure 9: The two assimilations with profile observations exhibit >1 line fit. Does this indicate assimilations with profile measurements tend to overestimate emissions or surface concentration?

Line 175: Please clarify the approach to project MOPITT CO AK and prior to the model resolution.

Line 195 and 196: the order of case 1 and 2 are opposite in the text and Figure 2.

Line 211: Did the authors consider a spin-down timeframe? For observations after the assimilation timeframe while can still reflect emission signals?

Line 320: Please clarify the definition of observation error variance, and if such error estimates have been incorporated into the assimilation system.

Line 331: Please clarify the meaning of x in eq 4.

Line 351: Experiment 2 and 5 are expected to assimilate similar information, with the major difference at 200 hPa, as indicated in Figure 5.

Line 420: The mean bias in text is 0.7 ppb, while 5.72 in Figure 9. The correlation is 0.65 in text, while 0.79 in Figure 9.

Line 465: Given the assimilation timeframe is 15 days, please clarify the approach to estimate annual emissions.

Section 7.4 How about temporal resolution? Can MOPITT optimize emissions up to daily scale based on the experiments here?

Technical corrections

Table 1: Missing the labels of TEMPO spectral ranges and potential chemical species for geostationary satellites.

Line 130: typo, an extra “;” after “CAM-chem”.

General Comments:

This study provides a comparative analysis by assimilating different MOPITT CO datasets to evaluate the advantages of assimilating multi-spectral satellite retrievals. This topic is important for better usage of MOPITT observations in the future. However, the manuscript focuses on the demonstration of assimilation results but lacks a further explanation for the possible causes of their large discrepancies, which reduces the importance of this manuscript. Furthermore, the content of this manuscript may need to be better organized. For example, the title of Section 4 is “Results”. The audience may assume it is the major discussion section, however, it is followed by Sections 5-7. In addition, there are lots of typo errors. Overall, this is a good study with potentially important impacts on the application of satellite data. However, a significant improvement may need to be made before the paper can be considered for publication.

Specific Comments:

• The authors raised three questions about the potential advantages of assimilating multispectral/joint retrievals. The developed and potential multispectral satellite retrievals are demonstrated in Table 1, including species of CO, O3, and CH4. More discussions are suggested to clarify the possible limitation of this analysis, i.e., whether the conclusion based on MOPITT CO is applicable to other instruments and species.
• The authors demonstrate dramatic discrepancies among different assimilations but lack a further explanation for the possible causes. For example, Figure 5: there are very large differences in the upper troposphere level. What are the possible causes and what are the possible impacts on the future assimilation of MOPITT CO data? Figure 10: Which assimilations have better performance in the upper troposphere, given their large discrepancies as shown in Figure 5? Section 6.1: The authors state “Assimilating profile products (Experiments (2) and (5)) tends to have a larger change to the emissions compared to only assimilating column products (Experiments (1), (3), and (4))”. However, the change by assimilating Profile JNT data is very small, which cannot be described as “tends to have a larger change”. The large discrepancy in Figure 13 by assimilating different datasets needs to be explained. Section 6.2: same as the above question, the change by assimilating Profile JNT data is almost ignorable, I don’t understand why it could lead to better agreement with observations.
• There are lots of typo errors. It seems the manuscript was not checked carefully. For example, Lines 421-423: “(1) Column JNT assimilation, (2) Profile JNT assimilation, (3) Column TIR assimilation, (4) Column TIR and column NIR assimilation, and (5) Profile TIR and column NIR assimilation”. Why are the five experiments listed here? Line 568 in the conclusion section, the authors state: “Results were not improved compared to WE-CAN because …” Because what?

Technical Comments:

• Line 32: “vertical localization” is frequently used in the manuscript, but the meaning is unclear.
• Table 1: what species are observed by the Geostationary satellites?
• Lines 155-156: “DART assimilates observations and produce the analysis, an ensemble of optimized initial conditions”; then in lines 164-166, it seems the emissions are also updated. What are the actual objectives to optimize? While references are provided, it would be better if the authors could clarify the assimilation methodology more clearly to facilitate the readers.
• Line 210: typo error: “run” at the beginning of this line.
• Line 213: it could be better to explain the cause of the higher fraction at the beginning of the assimilation here.
• The repeated usage of full experiment names, e.g., “Experiment (2) Profile JNT assimilation and Experiment (5) Profile TIR and column NIR assimilation” made the discussion hard to read. The authors may consider abbreviated names.
• The Conclusion section needs better organization. It is not good to list 9 main findings in the Conclusion.