Estimating forest height from InSAR and spaceborne lidar data over large areas are challenging but meaning work. However, the presentation of this manuscript makes it even more challenging to understand than it should be. Here are some comments that may be helpful:

1. Line 50: repeat "sensitive to". 

2. Line 70:use footprint instead of point. Point can be confused by lidar point cloud. 

3. Figure 2 and related text:Why not using Landsat/Sentinel based disturbance detection results, directly?  

4. Line 235 and equations: what does the a on the left mean? looks very similar to a. Suggest changing to other symbol. 

Also, where is hv(t2) ?

5. Line 245: It seems to be a typical tree height based allometric equation. But the parameters would vary a lot among tree species, and the forest age, determined by both t1, and t2. How were these uncertainties addressed?

6. Line 250: What is the since model for? Whether I missed it, or it failed to be introduced clearly. But it seems to be a very important one. Not clear how the values in the y axis of Figure 4 were calculated?

7. Figure 3: Again, as shown in Fig 3b, the growth rate varied a lot among site (species, age, and site condition). Also in Fig 3a, it should be a combined results of many different growth rates. These results further demonstrate it is unreasonable to apply a global model for the entire regions, even just for the New England region.  

8. 2.2.3: Oops, I got lost after the sinc model. Sorry. 

9. Figure 9: The flowchart definitely should come first, as the Figure 1 or 2. Also, make it a more general and easy to understand for general readers.

10. Figure 10:Labels on the color bar are too small to read. I would also suggest zoom into a few sub-figures of the study areas to show more details.  

11. Section 3: I would suggest put these parts before the methods. 

12. Table 2 and 3: Please add hom many plot or how large is the validate site in the table2 and 3. 

13. Fig 17,19,21,22,24,27,29, and so on. These figures are just too small to read clearly not to mention compare them. A good comparison should map the difference between the estimated and ground truth (ALS ERH98? maybe in Figure 29). 

14. Conclusions: I would suggest have a longer discussion and a short conclusions in separate sections. 

This paper presents a Radar/LiDAR fusion approach that create large-scale mosaics of forest stand height. After carefully reviewing the manuscript, I found this work presents some interesting results. However, there are some concerns needed to be addressed and clarified for improving the manuscript.

Major concerns.

1. The method of this work is based on temporal decorrelation modeling, but the Introduction part only presents and cites the author's previous work, without mentioning other well-known temporal decorrelation models. The introduction needs to be further improved.

2. Line 180: How is the size of the local window determined? The spatial density of GEDI is uneven, so why not use an adaptively varying window? Moreover, the local modeling approach is very similar to the work by Hu et al. (https://doi.org/10.3390/rs16071155), and they used an automatically varying window. It is recommended to add a citation to help readers better understand. In addition, using distance-based weighting does not seem to align well with the rapidly changed forest scenario. This should be further described and discussed.

3. The author emphasizes using the backscatter coefficient to estimate low forests, but there is no physical explanation for how the 10-m threshold is determined. Additionally, forest height below 10m can undergo significant changes, such as in young forests and shrub. When using ALOS data with a significant time difference for inversion, how is the height variation of short forests taken into account?

4. The content and structure of Section 4 are very redundant, with many figures and tables conveying the same information. Figures 26 and 31 are completely redundant; they have already appeared earlier in the manuscript, so why show them again? Additionally, the content of Tables 4 and 5 is the same as what is shown in the figures 26 and 31? Also, note that the accuracy metrics in Figure 26(b) are different from those in Table 4, please make the correction. In summary, the section 4 needs major adjustments and improvements.

5. The innovation of this manuscript lies in local modeling, and it is recommended to provide the results of global modeling for comparative analysis to highlight the improvement effect of the method.

Minor concerns.

1. Line 88: The wavelength of TanDEM-X is not ~0.01m, please check for updates.

2. Line 90: Are “these methods” referring to the TanDEM-X methods mentioned above? The method proposed in this paper may not necessarily outperform TanDEM-X. For example, the latest work by Qi et al (2025) adopts a strategy that is essentially similar to that of this manuscript.

3. I don't understand why the author chose this color scheme for Figures 13 and 14, and there is a lack of corresponding explanation.

4. Several global forest height products have been generated by combining GEDI and multi-source remote sensing data (Potapov et al, Lang et al.). The authors mentioned the limitations of these products in the introduction, and I suggest that the authors compare with these public products to highlight the performance of the proposed method and results.

5. There are several typos in the current manuscript. For example, Line 235: This finding i based on…; Line 445 Table3: left column, etc. Please proofread the manuscript carefully.