This is a review of the manuscript titled “A multi-year global methane data set obtained by merging observations from TROPOMI and IASI” submitted to ESSD by Shahzadi et al. The authors use the algorithms proposed by Schneider et al. to produce a daily, global, long-term methane dataset from TROPOMI and IASI satellite retrievals. Overall, this work continues their previous studies, fits the scope of the journal, and is timely because methane has recently risen sharply and exerts a strong warming influence on the climate. Therefore, I recommend publication in ESSD after the authors address several issues related to key technical validation details and the presentation of results.

Main comments:

1.    Although the authors have assessed the dataset accuracy using TM5 outputs and EDGAR emissions, ground measurements were not used for a more independent validation. I suggest a qualified comparison of the dataset with in situ/ground based results (e.g., TCCON).

2.    The manuscript states that IASI is “adjusted to TROPOMI a priori (TM5)”, but it is not explicit whether this is (a) a re retrieval with a different a priori, (b) an additive/multiplicative post processing correction, or (c) a linearization/offset mapping. The precise operation matters because it affects bias propagation and degrees of freedom.

3.    The Kalman update uses instrument posterior covariances as input. It is unclear how those covariances were pretreated (regularized, truncated, or inflated), whether they include forward-model uncertainty, and how numerical inversion (or pseudo-inversion) is handled for large matrices.

4.    Displacement errors are crucial for tropospheric products and are only described by reference to prior work. The current manuscript lacks the exact formulas, parameter choices (time/space scales), and tests showing sensitivity to those choices.

5.    The manuscript notes lower DOFS and increased noise in high-altitude areas, but does not quantify the implications for users (e.g., systematic smoothing, negative bias, or overconfidence).

Minor / Technical issues

1.    Add a detailed NetCDF variable list (names, units, dimensions), an explanation of averaging kernel format/shape, and a short example (Python/xarray snippet) showing how to read XCH4, troXCH4, the averaging kernel, uncertainty, and the quality bitmask. Provide a recommended QC selection for typical use.

2.    The selection of matching windows (50 km / 6 h, then normalized distance minimization) should be justified quantitatively. If possible, add sensitivity tests for different spatial/time windows (e.g., 25 km / 3 h, 75 km / 8 h) to show tradeoffs between coverage and displacement error.

3.    In the Methods section, define each symbol and its dimension immediately below the equation. For Eq. (4), (5), (6), and (9), clarify matrix sizes and transpose conventions, and explicitly state when operations are performed in log space versus linear space (you mention omitting log-transform details for brevity, please explicitly state in Methods whether final arrays are reported on the linear or log scale).

4.    For Fig. 10, ensure aerosol/snow masking and other flags are clearly labeled in the panels or legend, not only in the title; please check and apply the same clarity to other figures as needed.

This paper presents a global multi-year methane data product obtained by synergetically merging TROPOMI (XCH₄) and IASI (profile) Level-2 observations via a geo-matching step and a computationally efficient Kalman filtering scheme. The merged product provides three diagnostics—XCH₄, utsXCH₄ (≈450 hPa to TOA), and troXCH₄ (surface to 450 hPa)—with the stated goal of enhancing lower-tropospheric sensitivity and mitigating contamination from strong CH₄ gradients near the tropopause. The dataset covers Jan 2018–Jun 2021 with ~289 M combined samples, and the authors discuss uncertainty components (noise, dislocation/mismatch), DOFS patterns, and practical quality filters.

The contribution is valuable and timely. However, the physical and methodological explanation of why and how the Kalman synergy yields a lower-tropospheric partial column that is robust to tropopause height variability is under-developed in the current draft. The paper needs (i) clearer exposition of the state vector, observation operators, averaging kernels, and Kalman gain in the synergy; (ii) explicit sensitivity demonstrations showing reduced influence of UT/LS variability on troXCH₄; and (iii) several additional validations/robustness checks.

1. The manuscript claims and qualitatively illustrates that troXCH₄ (surface–450 hPa) is below the tropopause and therefore “independent from the strong CH₄ signals introduced by the location of the tropopause.” This is a central selling point and should be demonstrated more rigorously：Include row-wise AKs (and cumulative contribution functions) for several regimes (low vs. high tropopause, ocean vs. high terrain, clean vs. dusty/snow scenes). Demonstrate that 𝐴 for the tro layer largely suppresses UT/LS influence, while the uts layer captures most tropopause-related variability. This will convert a qualitative assertion into a physical demonstration.
2. Justify the 450 hPa boundary physically (global climatology of tropopause heights, retrieval DOFS distribution, and typical IASI vertical sensitivity). Provide a sensitivity check with an alternative boundary (e.g., 500 hPa) to show that conclusions are not fine-tuned to a single threshold. (You note that even extreme tropopauses at 300–400 hPa leave troXCH₄ below the tropopause; quantify this across latitudes and seasons.)
3. Add a controlled sensitivity test: perturb the a priori UT/LS by ±(50–100) ppb and show the response in XCH₄, utsXCH₄, and troXCH₄. I expect utsXCH₄ and XCH₄ to respond strongly, while troXCH₄ remains comparatively stable if the mechanism holds. Your Madrid/Iberia analysis already hints at this separation (XCH₄ shows superposition; troXCH₄ tracks near-surface seasonality/emissions), but a targeted experiment would make the case bullet-proof.
4. Provide a workflow schematic (geo-match constraints in space/time/surface pressure; selection of the “best match”; then Kalman update; then layer integration). Summarize the role of the surface-pressure proximity filter in reducing representativeness error before the merge. (Readers will appreciate why Δpsfc and distance/time windows matter for the dislocation kernel.)
5. Interpret the covariance terms physically: when and why does dislocation error rise (latitudinal gradient of temporal mismatch; high terrain where the tro layer is shallow), and how this interacts with the first kilometer above ground (largest representativeness uncertainty). The text mentions these patterns; add a concise, quantitative example (e.g., Himalaya vs. adjacent lowlands).
6. Clarify how the DOFS patterns co-vary with the noise maps (you note this relationship; add a 2-D density scatter to quantify correlation). Also, discuss the implications for regional comparability across seasons/latitudes.
7. Because troXCH₄ uncertainty grows over high terrain where the partial layer is shallow, add a topography-stratified analysis showing error growth and AK distortions with decreasing surface pressure, and recommend region-specific usage notes (e.g., Himalaya/Andes filters or uncertainty inflation).
8. A brief bias analysis for troXCH₄ XCH₄ against independent references (even if indirect) would greatly increase confidence.