Downscaling of SMAP Soil Moisture Data by Using a Deep Belief Network

Round 1

Reviewer 1 Report

This manuscript is appropriate for the remote sensing journal. This manuscript describes a study that involves downscaling of SMAP soil moisture using Deep Belief Network. The authors first identified the factors or predictors that are significantly corelated to soil surface moisture and secondly, they trained the DBN model. The model was then validated and compared with random forest model.  The developed DBN model gave satisfactory results and compared to random forest the DBN model results were better. This paper can be considered for publication after these revisions.

·       Make corrections in flow chart
Figure 2 – “Resempled” must be spelled as “resampled”

“ Wate mask” should be corrected to “water mask”

“Modelling” should be spelled as “Modeling”

·         It would be helpful for the readers if the authors can briefly explain how the low-resolution predictors were resampled?

·         Every downscaling method has its limitations or disadvantages. It will be helpful for the readers if the authors can elaborate with limitations associated with this approach.

·         Did authors encounter any missing data and how was it addressed?

·        Please read the manuscript carefully and correct the punctuation marks and grammar.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.Thanks!

Author Response File: Author Response.docx

Reviewer 2 Report

Downscaling coarse resolution soil moisture products into finer resolution has been serving as an effective way to get mid-high-resolution products for precision scale applications. Thus, the topic of this paper falls into the scope of remote sensing, and it may interest the readers of the journal. On the other hand, machine learning (especially deep learning) has been widely applied in remote sensing image processing, parameter estimation, scaling issue, etc. This paper utilizes DBN to downscale SMAP products, which is interesting and attractive. Here are some comments and suggestions for authors.

Major

1.        The English grammar has to be improved to make easy to read.

2.        The review of the progress in soil moisture downscaling can be improved and enriched.

3.        Necessary processing steps of the remote sensing and in-situ data can be briefly introduced.

 Minor

1.        SM “product” and “products” have been alternatively appeared. Please use the correct words according to the context.

2.        The definition of AMSR-E is not corrected, E means that Earth Observation System.

3.        The citation of references 11-22 and 34-36 can be distributed into each of the listed methods.

4.        Error metrics in Table 5 can be merged into Figure 6, no need to list as a table separately.

5.        Wrong x-tick-label in Figure 10b.

Author Response

Please see the attachment. Thanks!

Author Response File: Author Response.docx

Reviewer 3 Report

General：

According to the spatial resolution of current soil moisture products quality, the authors use a deep belief network-based method to downscale SMAP product. It’s happy to see downscaled 1 km result is validated by observation data. At the same time, the performance of the SMD-DBN method is compared with another commonly used machine learning model-random forest. The results show that the accuracy of SMD-DBN is higher than that of the SMD-RF. The authors also mentioned this method can be used to AMSR downscaling.

The downscaling method provided in this study provides a framework to support future hydrological modeling, regional drought monitoring, and agricultural research in the future. On the whole, the article is relatively successful. I suggest that the author carefully think about the following details and revise them before publishing.

Specific：

1.      L55, how do you think about “machine learning”? for physical process, machine learning is just a tool.

2.      L85, in study area section, can you add reference for climate back ground? Such as maximum inter-annual temperature of about 36 °C, a minimum of about -36°C. The average annual rainfall and evaporation is about 375mm and 1188mm. Where do you get this meteorological information?

3.      L97, please add scale in Fig 1.

4.      For remote sensing data, please add downloading website. Such as MODIS.

5.      In Fig 7,b,d,f,h show white part, what’s that meaning?

Author Response

Please see the attachment. Thanks!

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors have made necessary corrections, and they are suggested to check the grammar of the text carefully.

Author Response

Response to Reviewer 2 Comments

Point 1: The authors have made necessary corrections, and they are suggested to check the grammar of the text carefully.

Response: We would like to thank you for the comments and advice. We apologize for the language problems in the original manuscript. The language presentation was improved with assistance from a native English speaker with appropriate research background. And the formatting of the manuscript is checked against the journal instructions again. Revised portions are marked in red in the paper. Please find attached our revised document. The file named “remotesensing-1938573-track change” is revised version with all the changes still visible. We hope it can meet the journal’s standard. Thanks again.

Author Response File: Author Response.docx