This study by Saito et al. provides a high-level overview and evaluation of the GOSAT-2 XCH₄ product and the fluxes inferred from its use. GOSAT-2 is shown to produce significantly more successful observations than GOSAT-1, owing in part to its intelligent pointing. GOSAT-2 and GOSAT-1 XCH4 observations, in aggregate, are consistent, though the GOSAT-1 bias correction appears to drive a latitude-dependent bias between the two instruments. The GOSAT-2 observations are ingested into a previously described inversion system, with the posterior methane fluxes being remarkably different than the prior spatially (South America, Arabian Peninsula) and temporally (Northern Hemisphere seasonality). Generally speaking, the inversion using GOSAT-2 observations is consistent with the inversion using the GOSAT-1 observations, and both differ from the inversion using in situ observations, particularly close to the Equator. This paper will serve as a nice reference for GOSAT-2, and I recommend its publication with minor revisions described below, all at the discretion of the authors.

Minor comments

While reading the methods, I was interested in 1) how the pre- and post-screening works, and 2) the spectral range used for retrievals. Just a sentence on each would be helpful.

Line 256:  you write “repeated soundings over the same location within short time intervals were counted only once per day.” Could you clarify what you mean by this and when repeat soundings would occur?

For lines 291–293, could you add a sentence explaining how the South Atlantic Anomaly would impact the retrievals, and only for GOSAT-2 (as well as a reference if possible)?

Technical corrections

Line 17: would water not be the most abundant reactive greenhouse gas?

Line 20: “force” to “forcer”

Line 23: remove “to guide”

Line 42: might Frankenberg et al. (2005) be a better reference here?

Figure 4: maybe a red-white-blue color map would be easier to read for the “diff” plot.

Review of “Satellite-based inversion of global methane fluxes: Capabilities and implications of GOSAT-2 measurements”
Saito et al., ESSD, essd-2026-171

Recommendation: Accept with Major revision

General assessment

This manuscript presents a clear and useful description of the GOSAT-2 Level 2 XCH₄ product and the new GOSAT-2 Level 4 CH₄ flux product. The topic is well suited to ESSD, and the dataset is potentially valuable for the methane-cycle and atmospheric-inversion communities. The manuscript is generally well organized, and I particularly appreciate the comparison among GOSAT-2-only, GOSAT-based, and surface/aircraft-constrained inversions. This comparison is a useful way to evaluate the contribution of GOSAT-2 observations to regional and global CH₄ flux estimates.

I recommend publication with major revision. My concerns are primarily about documentation, uncertainty characterization, and interpretation rather than about the fundamental value of the dataset. In particular, because this is an ESSD data paper, the manuscript should provide enough information for readers and future users to understand the effective quality, uncertainty, limitations, and appropriate use of the Level 4 product without relying too heavily on external or forthcoming technical documents. I consider the manuscript suitable for publication after these issues are addressed.

Major comments

Provide stronger uncertainty and information-content characterization of the Level 4 flux product

The manuscript presents posterior CH₄ fluxes and posterior-minus-prior differences, but it does not provide enough quantitative information for users to assess where the G2L4 product is strongly constrained by GOSAT-2 observations and where it remains prior-dominated. This is particularly important because the manuscript itself notes that large posterior adjustments do not necessarily imply strong observational constraint; they may also arise in regions with weak sampling, persistent cloud cover, retrieval bias, or large prior uncertainty.

I ask the authors to add quantitative diagnostics of information content or uncertainty for the G2L4 product. Ideally, this would include posterior uncertainty, posterior/prior uncertainty reduction, degrees of freedom for signal, aggregated averaging kernels, or an equivalent observation-impact (or flux inversion performance) diagnostic. However, I recognize that full-resolution posterior covariance diagnostics may be difficult for a global 1° monthly 4D-Var system. Therefore, regionally aggregated diagnostics, e.g., over the 42 land regions used later in the manuscript and summarized by year or season, would be sufficient and very useful. 

If formal uncertainty or DFS diagnostics are not available, the manuscript should state this clearly and provide an alternative diagnostic or a more explicit qualitative description of which regions are likely to be observation-dominated versus prior-dominated.

Justify the use of a uniform 20 ppb observation prior error covariance

Section 2.3.4 states that each assimilated GOSAT-2 XCH₄ retrieval is assigned a uniform 20 ppb observation error. This is a consequential assumption because it controls the relative weighting of observations in the inversion. However, retrieval quality likely varies with local footprint characteristics.

Please provide a clearer justification for the 20 ppb value. Is this intended to represent an average error, a conservative upper bound, or a value inherited from a previous inversion configuration? If possible, please report G2L2 retrieval-error statistics, even in a compact form, such as a histogram or statistics stratified by land/ocean, latitude band, or surface type. If the authors retain the uniform value, please briefly discuss how this assumption may affect regional flux estimates and whether regions with poorer retrieval quality could be over-weighted.

Make the inversion methodology more self-contained

The manuscript refers readers to Niwa et al. and the G2L4 ATBD for several important details of the inversion system. While this is traceable, the present manuscript should include enough methodological detail for readers to evaluate the main assumptions behind the released data product.

Please add a compact, self-contained description of the inversion cost function, control vector, observation term, prior term, and prior covariance construction. In particular, please state the spatial and temporal correlation assumptions or localization scales used in the prior covariance matrix. Section 2.3.4 states that prior covariances for rice paddies, wetlands, and soil oxidation are derived from a 120-year VISIT ensemble with spatial localization, but the manuscript does not state the localization scale or whether month-to-month temporal correlations are imposed. These details are needed for readers to understand the effective resolution and degree of smoothing of the nominal 1° monthly product.

Address the bias-correction asymmetry between the G2L2 and G1L2 comparisons

A central comparison in the manuscript is between the G2L4 inversion and an analogous inversion using G1L2. However, the G2L2 product used for G2L4 is not bias-corrected, whereas the G1L2 V03.05 product used for comparison is bias-corrected. This asymmetry complicates interpretation of regional differences between the two satellite-based inversions.

This issue is especially important because the manuscript itself shows that the latitudinal structure of G2L2–G1L2 differences is sensitive to the presence or absence of the G1L2 bias correction; the latitudinal trend largely disappears when uncorrected G1L2 V03.00 is used. Therefore, regional differences between the G2L4 and G1L2-based inversions cannot be attributed solely to differences in GOSAT versus GOSAT-2 sampling, coverage, or retrieval characteristics.

Ideally, the authors would repeat the G1L2-based inversion using uncorrected G1L2 V03.00 for a cleaner like-for-like comparison. If this is not feasible, please explicitly qualify the interpretation wherever regional differences between G2L4 and the G1L2-based inversion are discussed. This would help readers understand how to interpret the differences in XCH4 and fluxes.

Clarify the design and interpretation of the three inversion experiments

The comparison among G2L4, the G1L2-based inversion, and the SURF+AIR-based inversion is one of the manuscript’s strengths. However, the rationale for the three-experiment design should be stated more explicitly. Please clarify what scientific question each comparison is intended to answer. For example: Does G2L4 reproduce the global CH₄ budget? Does the choice of satellite product alter regional flux attribution? Where do satellite-only constraints diverge from surface/aircraft constraints?

In addition, please clearly identify any differences among the experiments that could affect interpretation. Assumptions, priors, and constraints of each flux inversion in this manuscript can be put in a table to provide readers a concise way to understand the different flux inversion assumptions. For example, the manuscript notes that some prior components, including biomass-burning emissions, differ between the G2L4 setup and the Niwa et al. SURF+AIR system. If priors differ, then regional discrepancies cannot be attributed cleanly to the observational dataset alone. This caveat should be stated clearly in Section 4.1 and wherever regional G2L4 versus SURF+AIR differences are interpreted. This request will greatly help future readers interpret how the dataset advances the methane flux inversions.

Add normalized coverage or retrieval-success diagnostics

Figures 1 and 2 show raw counts of successful XCH₄ retrievals, which are useful. However, raw counts conflate several effects: orbital sampling, targeting strategy, attempted soundings, cloud screening, surface type, and retrieval success. A normalized metric would help users distinguish regions where few observations were attempted from regions where observations were attempted but frequently failed screening.

Please consider adding a figure or panel showing retrieval-success fraction by latitude band or region, ideally normalized by the number of attempted soundings or by the orbit-geometry-limited maximum. If attempted-sounding counts are not readily available, please discuss this limitation and consider adding an alternative normalized coverage metric.

Be more cautious in interpreting regional and sector-dominated flux adjustments

The manuscript includes a useful discussion of posterior adjustments in anthropogenic-source-dominated regions. However, the inversion optimizes net surface CH₄ flux and does not directly isolate sector-specific emissions. Therefore, sectoral interpretation should be framed carefully.

Please clarify that the sectoral interpretation is based on prior sectoral dominance or regional source composition, rather than on a sector-resolved inversion. For regional adjustments such as the large negative adjustment in East Asia, please provide posterior uncertainty estimates if available. If formal posterior uncertainties are not available, please state this clearly and frame these adjustments as qualitative or diagnostic rather than as robust sector-specific corrections.

Improve data-product documentation and reusability

For reuse, the manuscript dataset should clearly document variable names, units, dimensions, time conventions, missing values, prior and posterior fields, posterior-minus-prior fields, uncertainty fields if available in tabular form added to the appendix. This should be added also added to the Zenodo archive with a README explaining which netcdf files correspond to which dataset and time period. The manuscript should also have recommended citation for the dataset as well.

Additional comments

Section 2.3.3: Please clarify how EDGAR v6.0 anthropogenic emissions are handled for 2019–2022. Are anthropogenic prior emissions held fixed at the final EDGAR year, extrapolated, or treated in some other way? This matters because any real anthropogenic emission trend not represented in the prior could appear as a posterior correction.

Please make figure captions more self-contained. For maps and regional comparisons of posterior–prior differences, please specify whether values are annual means, annual totals, monthly means, or flux densities, and state how missing data or masks are handled.

The manuscript has some awkward phrasing. For example, phrases such as “to guide to constrain atmospheric concentrations” and “expanded observational capacities of GOSAT-2 retrieval to its predecessor” should be revised for clarity.

Summary

This is a valuable manuscript describing a potentially important GOSAT-2 methane flux data product. I support publication after revision. The main changes needed are to make the inversion assumptions more transparent, better characterize uncertainty and observational constraint, address bias-correction and prior-related confounds in the intercomparison, and improve documentation for future users of the dataset.