General comment:

This study assimilates the SMAP surface soil moisture product into the Noah-MP land surface model over the eastern continental U.S. for 2016 and evaluates its performance using multiple reference datasets, including model-driven products and in situ observations. The authors report strong overall performance of SMAP assimilation while noting that regions with higher errors tend to be associated clay soil types, highlighting directions for future improvements. The manuscript is well-structured and clearly written. However, there are two main weaknesses: 1) a lack of novelty in advancing soil moisture data assimilation research and 2) a lack of consideration on the spatial scale mismatch between model-derived products and in-situ observations. Please see my detailed comments below and I hope those will help enhance the manuscript.

Specific comments:

1. Introduction: The literature review primarily focuses on studies that assimilate SMAP into Noah-MP. However, it would be beneficial to broaden the discussion to include studies that 1) assimilate different soil moisture products and 2) employ DA with other land surface models, providing a more comprehensive overview of the benefits, limitations, and current challenges. And then clearly state the major contribution of this study to the soil moisture DA field. For instance, SMAP has a relatively coarse resolution, whereas other satellite-based soil moisture products such as Sentinel-based ones offer finer spatial scale. Why the authors choose to use SMAP for a 1-km assimilation set up? The introduction could be benefited from discussing how assimilating higher-resolution soil moisture products affects soil moisture estimation and how the assimilation of coarser SMAP data compares to these alternatives. Overall, the study’s novelty and unique contributions relative to existing literature should be explicitly articulated.
2. Analysis domain and period:Since SMAP has been available from March 2015 onward, why was data assimilation performed for only a single year? A longer assimilation period could provide insights into the model’s ability to represent interannual variability and extreme events such as drought. Since the authors did have a section discussing drought, I think it is not sufficient to discuss drought without referring to a long-term reference period. Please see my comment #12 in more detail.
3. L167-168: Is there a reference to support this claim, in particular reference that supports the evidence of Stage IV QPE being better than NLDAS-2 for the study period.
4. Section 3.3: Some methodological details are unclear or missing: 1) Is there any spin-up process for the OL simulation, and does it reach equilibrium before applying the restart file for DA? 2) What is the reference period used for the monthly CDF matching? 3) Why use CDF-matching strategy? Does the CDF-matching strategy risk distorting real signals in SMAP data by adjusting them to fit the model-derived distribution? How do the CDF profiles compare to model-based CDF before and after CDF-matching? Some studies favor anomaly scaling over CDF matching to better preserve variability in satellite-based observations – additional analysis may be required to justify the choice of CDF matching.
5. Evaluation: does the authors consider the correlation for anomalies by removing the monthly mean? Analyzing anomaly correlations could provide additional insight into performance, as seasonal cycles may dominate raw correlations. Besides, since monthly CDF matching is used, it might be also beneficial to compare OL against the reference datasets as correlation may be largely dominated by the seasonal variation from OL. Comparing OL vs. DA would help distinguish the improvement from DA.
6. L281-284 "...there are grids received zero updates, especially in the eastern part of the domain..." This needs clarification. Is this due to missing raw data, quality control procedures in SMAPDA, or another factor?
7. L 285-291: It would be beneficial if the authors can elaborate the discussion on what might be causing the seasonal patterns of the increments. For instance, what processes might be captured by SMAP but not represented in the model? Does SMAP assimilation primarily correct model soil moisture errors originating from forcing uncertainties, or does it capture external inflences such as irrigation, which are absent in the model? A deeper discussion of these factors with supported reference would strengthen the manuscript.
8. Section 4.2: The authors acknowledge to some extent that scale mismatch between SMAP (9 km resolution) and in situ observations could degrade evaluation metrics. Would it be possible to aggregate the relatively densely distributed OKMet or SCAN data to a resolution comparable to the model-based reference products and raw SMAP and reassess the metrics? For instance, how does the metrics change if evaluating at the ERA5-land or raw SMAP resolution? Addressing the scale mismatch explicitly would improve the fairness of the comparison and strengthen the conclusions.
9. Figure 5 and 6: It is hard to visually distinguish the difference across the comparisons. Please consider keep the RMSE and Bias plot for SMAPDA vs. USCRN to illustrate the spatial pattern of the metrics while plot the RMSE and Bias differences for other four comparisons against that of SMAPDA vs. USCRN. This would help the audience better see locations where other datasets perform better or worse compared to SMAPDA. The same suggestions apply to the supplementary figures.
10. Figure 9: The smaller STD in SMAP AM could be a consequence of the coarse spatial resolution. To ensure a fair comparison, it is better to scale OKMet to a comparable resolution before computing STD. Alternatively, the manuscript could discuss about the sensitivity of STD estimates at different spatial scales.
11. When assessing ST, LHF, and SHF, how does OL compare to SMAPDA? Since these variables are indirectly influenced by SMAPDA, it would be informative to first demonstrate the OL simulation performance to better understand the extent to which SMAP assimilation modifies these variables.
12. Section 4.2.4: A single year of data is insufficient to robustly assess drought conditions, as drought classification typically relies on long-term climatological baselines. Do the authors have a longer-term OL simulation with sufficient spin-up? Instead of relying solely on trends within a year, it would be more appropriate to rank the 2016 conditions relative to a multi-year climatology and analyze its deviation from historical conditions.

This study assimilates SMAP (9 km) surface soil moisture estimates to generate 1 km soil moisture, fluxes (sensible, latent), and temperature estimates for a region (Central, SE) of CONUS that's of critical interest (high density of in situ networks) for drought monitoring. Other datasets (ERA5 [reanalysis], GLASS [ML], GLEAM [model], SMAP [remotely sensed]) are compared with the available in situ records for soil moisture or other land-atmosphere variables as available. The 1 km SMAPDA product shows relatively good performance, and reasons are suggest further improvement for soil moisture modeling. This study helps advance downscaling/modeling soil moisture with land surface models (specifically Noah-MP). However, as the first reviewer has mentioned, further discussion is required to help frame the significance of this experiment. The limited data generated also needs to be considered (1 year, 2016). How does comparing a 1 km SMAP DA-modeled soil moisture product against other soil moisture products (modeled or otherwise) for this 1 year contribute to either the LSM, soil moisture/climate community, or other stakeholders?

Major comments:

1. Why was SMAP 9 km used to downscale to a 1 km resolution? SMAP-based 1 km soil moisture datasets exist (i.e., the NSIDC-SMAP 1 km product by Fang et al., 2022) that is available for this study period and area. Sentinel-1 based soil moisture would also be good to consider/compare as well (Brocca et al., 2024).

2. Why is only SMAP AM used for the assimilation? The afternoon product is mentioned, but I did not see any further justification for excluding it. Also, note that the SMAP L3 product is not exactly daily due to missing coverage. Was any gap-filling used prior to the data assimilation?

3. The resolution of each respective product appears to be an important point for explaining the pattern in performance (e.g., GLASS & SMAPDA vs. SPL3SMP_E/ERA5-Land/GLEAM). Why didn't the authors attempt a higher resolution grid (e.g., 500 m) for the forcing? Given the Rouf et al. (2021) paper investigated up to that resolution with Noah-MP, such a study may be more impactful than the current one presented. Their study also cite increasing accuracy with increased forcing resolution.

4. Why was only the upper soil moisture compared? I understand SMAP is the limiting factor here, but I'd argue the benefit of a (DA) model is to investigate the variables (e.g., root zone soil moisture) not directly sensed by the satellite. Given the comparison with ERA5-Land, soil moisture at lower depths would be helpful for comparison/validation as well.

5. An open-loop simulation would be important for a base comparison. How do you justify not running OL simulations and comparing them against the DA outputs? Furthermore, more details on the modeling set-up could be provided. How much spin-up time was used? Although the purpose for analyzing data from 2016 is given (major drought year), either more work or reasoning needs to be done for limiting the study period to one year. E.g., how do you know your results are representative of the datasets themselves, and not just valid for times of drought?

Minor comments:

6. The number of stations should be explicitly described, and/or mention the OK (the state) in Figure 1 when describing the ARM SGP sites. The white circles were a little difficult to identify at first. The number of stations for validation would also help contextualize Figure 4.

7. Line 45 - is (O (10 km)) a typo?

8. Line 320, Figure 4 - GLASS is stated to have more off-diagonal terms, but the concentration and shape of off-diagonal terms in the SMAPDA plot make that statement appear moot. The total RMSE for GLASS is also slightly lower than the SMAPDA product.

9. Line 482 - Sites 6,8,14 also show high bias during some months but are more sandy soils. What do you attribute the cause to be? Could it be a question of sampling? (e.g., less clay sampled available, whereas more sandy time series exist)

10. Line 444 - The soil moisture bias doesn't appear to be that great compared to the insitu data. And LHF is only marginally larger for SMAPDA than the observed during the summer. The differences in the summer SHFs also show that they are not within error/standard deviations. Are there other causes you could attribute this discrepancy to?