Comparison of QRNN and QRF Models in Forest Biomass Estimation Based on the Screening of VIs Using an Equidistant Quantile Method

Round 1

Reviewer 1 Report

Comparison the performance of QRNN and QRF models based on the VIs screening by equidistant quantile method.

This research investigates the potential correlation between filtered vegetation index and forest aboveground biomass (AGB) using traditional variable screening methods. The research specifically examines Pinus densata forests in Shangri-La, employing 31 variables to establish quantile regression models for AGB at 19 quantiles. The study identifies key variables related to biomass and constructs QRNN and QRF models. Notably, it identifies that certain bands exhibit significant relationships with AGB at specific quantiles, emphasizing the importance of band selection. The study concludes that a vegetation index involving band 2 and SWIR is more suitable for this region for Pinus densata. The QRNN and QRF models demonstrate their best fit at the 0.5 quantiles, with respective R² values of 0.68 and 0.71, effectively reducing underestimation and overestimation. Overall, the research contributes valuable insights into variable screening methods, enhancing estimation accuracy and mitigating underestimation and overestimation issues.

Page 19, line 187, hall et al. (1995) capitalization issue.

Page 19, “….nor it is suitable for forest land with complex structure (Sader et al. 1989). What makes a complex structure?

Page 21, line 240 Zhang’s stud??

Page 21, line 241 But higher than the studies mentioned above.??

Data is well analyzed, but explanation is little convoluted in some places making it difficult to grasp the concept. Refining languages conscientiously and checking references would improve the quality of paper. Both abstract and conclusion may be rewritten to make them smooth flow.  Other than that this paper has some merits and worth publishing.

Needs nicely polishing the language. 

Author Response

Thanks for your warm comments, we've revised the manuscript according to your precious advice, please find the details in the attached file.

Author Response File: Author Response.pdf

Reviewer 2 Report

I have reviewed the paper and it is fine in my view for publishing with some review of the English. I also have some comments for the authors.

The research topic is interesting, but its still far to demonstrate concretely that it increases the accuracy of the biomass predictions and some concerns remain on the superiority of this aproach to SAR data and LIDAR data and the scalability of  suchapproach to large scales.

Author Response

Thanks for your warm comments, we've revised the manuscript according to your precious advice, please find the details in the attached file.

Author Response File: Author Response.pdf

Reviewer 3 Report

Reviewer response to the authors.

This manuscript described a study evaluating the predictive models for the forest AGB estimation in Yunnan province using Landsat 8 and ground data. The results were good and showed the insight models. However, there are several shortcomings needing attention:  The title is needed to improve. 

1.     I would recommend the authors improve the abstract, especially in the method parts

2.     In Figure 1, you have to provide more comprehensive such as VIs calculation and spectral extraction.  It should be clear to the readers.  

3.     The authors have to improve the study area map. For example, the background is unclear to see details, and a scale bar and arrow should be provided for countries and province maps. In addition, a description of Figure 2 should be provided with the background and plot sampling.            

4.     I would suggest adding more information about the inventory data. How did you separate the training and validation for your study?

5.     Why did you use the L8 images in December?

6.     Why did you use 25 VIs for the models? The authors need to provide the reasons. 

7.     How did you use the VIs and spectral bands for constructing the predictive models?

8.     I recommend you improve the model implementation, such as parameter setting and selected features. This part (2.3 models) is very unclear to me, and it is an important part of your work.

9.     Why did you make the interval values (2.4 Model evaluation)?

10.   In Figure 7, the authors must show the map with singleband pseudocolor. Also, The map should be improved and updated.

11.  In the discussion part, what do you think if we use Sentinel-1 and Sentinel-2 in your models? What do you think about applying other machine learning models?

12.  In conclusion, I suggest you add the benefits of your work for the future study. 

Specific Comments/Suggestions

1.     Please check the format.

2.     Pease check the references.

3.     In the study area, “km2” or “km²”    

4.     All figures are needed to improve the resolution.

5.     Please provide the references for VIs.

The English language is required to improve.

Author Response

Thanks for your warm comments, we've revised the manuscript according to your precious advice, please find the details in the attached file.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

My comments have comprehensively addressed

Author Response

Thanks for spending time on review our article, thanks again.