This study provided an insightful analysis of the pros and cons of multi-sensor soil moisture data assimilation source on a global and regional basis. Using multi-sensor soil moisture remote sensing seems like a valuable and generally more robust approach based on the results provided in the study. The authors also clearly outlined the limitations of the multi-sensor approach and ways to improve it in the future. The findings are interesting because they avoid relying on blended soil moisture datasets, also allowing for a more precise evaluation of each individual soil moisture product.

I found the paper well-structured and informative, though the frequent use of acronyms made it a bit harder to follow at times. That said, it's understandable given the models and datasets that were used. My review focuses primarily on how certain results are framed and interpreted in relation to the study’s overarching theme. I also included a few line-specific comments aimed at improving the clarity of specific passages.

General comments

1. The study presents the robustness of SG_AT and SG_SP in the TCA section, but MT_ATSP was excluded due to the restrictions of TCA. However, the added value of multi-sensor DA being the main theme of the study, this makes it difficult to fully interpret the TCA results in the broader context of the study. The TCA results, particularly in Figure 3, are valuable and suggest that SG_SP is generally the more robust option for single-sensor DA. Still, the authors should clarify how these results relate to the overall narrative of the study, given that MT_ATSP is the preferred DA method despite its absence from the TCA.
2. The effect of SM DA on precipitation forecast skill is not clearly explained, which makes it difficult to understand the reasoning behind evaluating precipitation forecast. To improve coherence, the authors should clarify how SM DA is expected to influence precipitation forecasts and why these metrics are relevant to assessing the performance of the assimilation methods.
3. There appears to be a slight disconnect between the presentation of results in Sections 7.1–7.2 and Section 7.3. While the earlier sections highlight the relative performance of SM DA methods compared to the control scenario, the precipitation results are presented in absolute terms, which makes it harder to assess the added value of each method. In Figure 7, the relative performance has to be inferred visually rather than being directly quantified. Is it more important here to show the absolute forecast metrics, or to compare them against the control baseline? Presenting ∆FB and ∆ETS scores could make these comparisons clearer and more consistent with the rest of the paper. Boxplots similar to those in Figure 3 might be a good way to present these results more intuitively.
4. In the conclusion, the authors should reiterate that a strength of the study is its use of multiple independent SM datasets, rather than relying on blended products, and briefly explain why it matters. 

Specific comments

Line 164: a proper minus sign should be used to prevent the line break with the following number (this applies to the instances afterwards as well).

Line 166: unclear what “and restart files at 0 h” means, could you clarify?

Line 179: what are the advantages of the KIM-LIS coupled system mentioned here?

Line 227: please state the specific DA assumption being cited.

Line 232: I think that the reasoning for using different bias correction methods for the two SM datasets should be clearly stated here.

Line 409: please change to “March 1^st 2022” or an equivalent format.

Line 458: I think the formulation for  should be presented the same way as equations 2 to 4 for consistency.

Lines 488-492: more information to justify the timestep of the LSM outputs that were used would improve clarity.

Line 616: “RMSD difference” is used here, but later, in tables 5 and 6,  is used. Please choose one for consistency.

Lines 628–632: the statement that MT_ATSP results in a more “balanced improvement” than single-sensor DA methods needs clarification. What does it mean in this context?

Line 713: I assume that the FB and ETS were computed for the 00 UTC cycles for all the days in July? The methodology to compute the precipitation amounts is a bit unclear.

Line 820: Please clarify if Figures S4 and S1c come from Kim et al. (2025) as well.

This manuscript evaluates the impact of assimilating ASCAT and SMAP soil moisture retrievals, both individually and simultaneously, within the KIM-LIS weakly coupled land-atmosphere data assimilation system. The aim is to assess the potential synergistic effects on global soil moisture analysis and numerical weather prediction skill. This work has significant prospects for application on related platforms and systems. The efforts made in this work are highly commendable. However, the validity of the evaluation methodology and the significance of the marginal improvements achieved are called into question. My suggestions are as follows:

Major comments:

1. The methodology used to quantify the improvement in accuracy from data assimilation (reported as 4.0% and 10.5% for ASCAT and SMAP, respectively) raises serious concerns. The approach of comparing TC-derived error estimates from two different triplets—specifically, comparing the error of CTL in the [AMSR2, SMOS, CTL] triplet with the error of the DA experiment (e.g., SG_AT) in the [AMSR2, SMOS, SG_AT] triplet—may not be fully justified. Since TC computes errors relative to the entire triplet in which a dataset is embedded, replacing one member (CTL with SG_AT) changes the reference framework and can lead to a re-balancing of the error estimates for all three components. This means the error estimates for CTL and the DA experiment may not be directly comparable across these two separate TC configurations. Therefore, the reported percentage improvements could be influenced by methodological artifacts rather than reflecting a true measure of performance.
2. A critical issue lies in the very limited to negligible improvement in the forecasts of key atmospheric variables. In some cases, negative skill increments are observed. These results significantly undermine the practical justification and operational feasibility of the proposed multi-sensor assimilation approach. Consequently, the study fails to provide readers with quantifiable and meaningfully positive conclusions regarding the benefits of simultaneously assimilating soil moisture retrievals for enhancing numerical weather prediction.
3. While the introduction outlines what the study does, it falls short in providing a comprehensive literature review and a compelling justification for why this study is necessary. Specifically, (1) the objective of enhancing soil moisture estimates by integrating satellite data is clear. however, the introduction lacks a critical discussion on the broader landscape of methodologies available to achieve this goal. Notably, it omits any mention of alternative approaches, such as statistical fusion techniques or the rapidly advancing field of machine learning, which have been extensively employed for soil moisture reconstruction, data fusion, and even forecasting. (2) The introduction describes the applicability of data assimilation in general (Lines 36-50), the use of ASCAT and SMAP in data assimilation (Lines 67-80), and the combination of active and passive sensors (Lines 81-93). Yet, it fails to clearly articulate the specific research gap and the novelty of this particular research. The reader is left wondering: What is the unique contribution of this work? Is it the use of the specific KIM-LIS coupled model platform? Is it the simultaneous assimilation of ASCAT and SMAP retrievals? If it is the former, the authors should more clearly articulate what makes the KIM-LIS platform itself a novel or particularly advantageous choice for this specific investigation, beyond merely being the system used. If the latter case, how does this approach differ from and improve upon previous studies that assimilate multi-source data?
4. The manuscript is overloaded with abbreviations, making it exceedingly difficult to read. This is particularly problematic in the Abstract and Figure Captions, where the pervasive use of undefined acronyms obscures the authors' intended message and hinders comprehension. The authors should prioritize clarity over brevity and significantly reduce the use of non-standard abbreviations to ensure their work is accessible to a broad audience.

Minor comments:

1. Line 1: The abstract lacks a clear statement of the research motivation. It should briefly highlight the importance of assimilating multi-sensor soil moisture data for improving numerical weather prediction to better contextualize the study for reader.
2. Lines 405-412: The choice to use GLDAS for LSM spin-up but ERA5 for model initialization may introduce inconsistencies. Although GLDAS is a land data assimilation system and ERA5 is reanalysis, both of them provide land surface data and atmospheric forcing data. The authors should address whether this discrepancy in forcing data sources could have impacted the results.
3. Lines 421-422: “The four experiments (i.e., CTL, SG_AT, SG_SP, and MT_ATSP) listed in Table 1”. The experiments are listed in Table 2.