Application of Gaofen-6 Images in the Downscaling of Land Surface Temperatures

Round 1

Reviewer 1 Report

Thank you

Author Response

We would like to express our great appreciation for your comments.

Reviewer 2 Report

After scrutinizing your (revised) manuscript, I found it publishable in the present form. There are but a few very small corrections necessary which I marked in the annotated PDF attached.

Comments for author File: Comments.pdf

Author Response

Thank you for the comments. Those comments are all valuable and very helpful for revising and improving our paper. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion were mainly marked with track changes in the paper. The responds are as following:

Point 1: After scrutinizing your (revised) manuscript, I found it publishable in the present form. There are but a few very small corrections necessary which I marked in the annotated PDF attached.

Response 1: The faults marked in the PDF attached have been revised. 

Reviewer 3 Report

This paper compares multiple methods for downscaling coarser resolution Landsat 8 LST images using higher resolution GF-6 solar band images, over the Ebinur Lake Watershed. Results indicate that the downscaling LST using NDVIRE2, RBI, NDSI, and NDWI with MIRF method performs the best. Many data, methods, and results are involved in this study. However, they are not well presented at the current stage.

1. The track change version of the manuscript is hard to read. A clean version needs to be provided in the future.
2. Abstract: (1) please define all terms before they are used, such as GF, MIRF, DNVI, RBI, DNSI, NDWI, … (2) Suggest using more specific terms. For example, use Landsat 8 retrieved LST (line 14) instead of LST.
3. Table 1: the information listed in this table are too general.  Please include which GF-6 WFV bands (and center wavelengths), Landsat 8 OLI and TIRS bands (and center wavelengths), which GF-6 & Landsat 8 image pairs were used in this study.
4. Table 2: What measurements are used from each station? LST is usually not directly measured, instead, they are converted from some other measured parameters such as upwelling longwave radiation and surface emissivity. How was the ground measured LST derived in this study?
5. Figure 2: Please revise the caption. What does each data point represent, individual pixel level DNVI or some averaged values?
6. Figure 4: please clarify which GF-6 bands were used to plot the three color composite images in the caption.
7. Figure 5: what’s the data source of the surface types? Were they generated in this study?
8. Figure 6: please clarify the LST images shown in this figure are retrieved using data observed by which satellite.
9. Section 2.2.3: Please give a brief background about the LST retrieval method used. Is this method work well for all land cover types? How are the parameters given in Equations 5-7 derived? Are these values site-specific?
10. Figures 7-9, 10-12, 13-15, and 16-18: there are too many images/plots in these figures. Suggest showing the representative cases (images/plots) only and summary the statistics for all cases using tables. In addition, figures 7-9 are too far away from the text.
11. Page 4, line 173: “the transit time”: do you mean the time for satellite observations?
12. Terms such as “remote sensing data”, “remote sensing images”, “remote sensing indices”, “remote sensing indicators”, and “remote sensing factors” appear frequently throughout the text, which make the paper hard to follow. Please be more specific when it is possible, for example, using terms such as Landsat 8 retrieved LST and GF-6 NDVI.
13. It will be more useful to provide information about how the downscaled 16 m LST images developed in this study will be used by the users.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

The manuscript has been improved significantly after the revision, and it is easier to read than before. Following are my comments:

1. Table 1: thanks for the improved the table. Please add the unit of spatial resolution. Are all GF-6 and Landsat-8 bands used in this study? Suggest only listing the bands that are actually used.
2. Lines 192 – 196: Equation 1 seems derived from Stefen-Boltzmann law, which describe the total energy radiated from all wavelengths. What is the spectral range of ground instrument that measures the downwelling and upwelling longwave radiation? What is the emissivity for each site? What is the source of emissivity data?
3. Section 2.2.3: which GF-6 and Landsat-8 thermal bands are used for the LST retrievals? References 42-44 seems for Landsat-8 LST only. What’s the performance of this LST retrieval method for GF-6? In addition, are the GF-6 and Landsat-8 LST bands cross-calibrated? What is the source of emissivity data for Equation 2?
4. Line 237: “the retrieval error of LST decreases by 1.5-3K”: do you mean the errors of LST is between 1.5 – 3K or decreased by 1.5 -3 K from some original values? 
5. Line 255: I guess T₀ is air temperature measured at 2m above the surface. Please clarify.
6. Line 314: please define RE1 and RE2 before they are used.
7. Lines 356 – 358 & Figure 6: it is not clear which LSTs are derived using GF-6 data and which LSTs are derived using Landsat-8 data.
8. What is the spatial resolution of the LST images shown in Figure 7?
9. Lines 336-346: (1) Equation 26 seems not complete. (2) RMSD (or RMSE) equation is well-establish. There is no need to show its equation in the paper.   
10. Remove the paragraphs from line 454 to 490, which just repeat the statistics summarized in Table 4.
11. Line 491: it is not clear what is meaning for “The evaluation results that the three groups of images show consistent regular (Table 4).” Please revise.
12. The English still needs to be further improved.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

1. The article is prepared carelessly and requires both linguistic and editorial correction. Many, but probably not all, faults of this type are indicated in the attached PDF file.
2. Descriptions of some stages of the methodology are omitted or are too superficial, and therefore it is difficult to assess the quality of the results obtained. For example, it is not known what criteria were used to select Test Area A (Figs. 5 and 6). It is not known on what basis the area was classified into the four land cover categories (vegetation, water, impermeable surfaces and barren soils) and the accuracy of this classification. Other remarks of this type are also placed next to the text in the PDF file.
3. However, the most important reservations concern the lack of independent data to validate the quality of the performed downscaling. In my opinion, there is no better method for the one that produces images with more small differences (details). The greater number of details does not mean that they represent the real variability of the LST. Treating the parameters R2 and RMSD as quality indicators in relation to the original (source) LST image is also in my opinion wrong. Given such criteria, the best downscaling method would be to re-sample from a sparse mesh to a finer mesh. 

Comments for author File: Comments.pdf

Author Response

Thank you for the comments. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion were mainly marked with track changes in the paper. The responds are as following:

Point 1: The article is prepared carelessly and requires both linguistic and editorial correction. Many, but probably not all, faults of this type are indicated in the attached PDF file.

Response 1: The faults indicated in the PDF file have been revised. The revised part has been edited by linguistic and editorial corrections.

Point 2: Descriptions of some stages of the methodology are omitted or are too superficial, and therefore it is difficult to assess the quality of the results obtained. For example, it is not known what criteria were used to select Test Area A (Figs. 5 and 6). It is not known on what basis the area was classified into the four land cover categories (vegetation, water, impermeable surfaces and barren soils) and the accuracy of this classification. Other remarks of this type are also placed next to the text in the PDF file.

Response 2: We have revised the description of the method. According to the field visit and the GF-6 image of the study area, underlying surface types of the study area mainly include vegetation, water bodies, impermeable surfaces, and barren soil.GF-6 image of the study area was used to classify the underlying surface into these four types, and the classification accuracy was 94.4% with support vector machines (SVM). We selected region A to show the details of the results, because it is located in the central area of the county seat, and which includes all four typical underlying surface types. Other remarks of this type in the PDF file have been all revised.

Point 3: However, the most important reservations concern the lack of independent data to validate the quality of the performed downscaling. In my opinion, there is no better method for the one that produces images with more small differences (details). The greater number of details does not mean that they represent the real variability of the LST. Treating the parameters R2 and RMSD as quality indicators in relation to the original (source) LST image is also in my opinion wrong. Given such criteria, the best downscaling method would be to re-sample from a sparse mesh to a finer mesh.

Response 3: Think about the suggestion carefully, we have revised the Materials and Methods and Results section. Consider that Landsat-8 TIRS images from USGS have been resampled to a resolution of 30 m, GF-6 and Landsat-8 TIRS images were first scaled up to 100 meters to obtain the remote sensing index and LST under the low resolution, and to establish downscaling model. GF-6 images with original resolution were used to construct the high resolution of the remote sensing index to downscale on the low-resolution LST, and 16m LST was obtained. Landsat-8 OLI and the original TIRS images before scale-up were used to retrieve the LST, and to evaluate the downscaling results of the resampling to 30 m. Direct verification of measurement site and cross validation of reference LST were used to evaluate the downscaled LST. So we used R² as the coefficient of determination between the original(the retrieved LST using Landsat-8 OLI and the original TIRS images) and downscaled images, and RMSD was used to test the difference between the original and downscaled LSTs, and a high R² and low RMSD indicates a satisfactory downscaling.

Author Response File: Author Response.docx

Reviewer 2 Report

This could be an interesting work if the following issues are properly addressed:

[1] It is not clear whether you used Landsat 8 to downscale GF data? So be specific and focused to clarify everything

[2] The intro section has used works that are pretty old. I would suggest to clearly lay down motivation of your work. The use of land surface temperature is highly concentrated on surface urban heat island computation than any other areas as it is a useful input. However, your intro part is really poor to demonstrate how this work would be useful. Furthermore, Landsat could be contaminated with cloud especially in tropical and sub-tropical countries. This is not noted at all. I therefore would suggest to review these works to better shape the novelty of this work (https://www.sciencedirect.com/science/article/pii/S0924271620302082; https://www.sciencedirect.com/science/article/pii/S2210670721002122;  https://www.sciencedirect.com/science/article/pii/S2210670721002900

[3] You did use three methods to downscale, this means there is little contribution to literature. I do therefore disagree with the statement in line 67-68. Hence you must show novelty and contribution of this work to literature clearly

[4] You generated ground LST with instruments. What instrument was used? How long they run for? What was their reading resolution? Was there any calibration done? Nothing noted. How did you choose sitting locations? Arbitrary?

[5] Table 3: how did you do normalisaiton? Though remote sensing community knows this well you cannot expect all readers may know this. So clarify

[6] Precision metrices – this heading is wrong.

[7] Fig. 3 – where did you get Landsat 9 LST product? Did not mention, how was it computed? Your method of LST is different from product LST? How error value in fig. 4b obtained?

[8] Need to show info re GF satellite in supp information

[9] fit lines are not well presented in all related figures, hence needing change

[10] Discussion section present but not well correlated with theory  

Author Response

Thank you for the comments. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion were mainly marked in red in the paper. The responds are as following:

Point 1: It is not clear whether you used Landsat 8 to downscale GF data? So be specific and focused to clarify everything

Response 1: We revised the data usage and experimental methods section to clarify the experimental process.

Point 2: The intro section has used works that are pretty old. I would suggest to clearly lay down motivation of your work. The use of land surface temperature is highly concentrated on surface urban heat island computation than any other areas as it is a useful input. However, your intro part is really poor to demonstrate how this work would be useful. Furthermore, Landsat could be contaminated with cloud especially in tropical and sub-tropical countries. This is not noted at all. I therefore would suggest to review these works to better shape the novelty of this work (https://www.sciencedirect.com/science/article/pii/S0924271620302082; https://www.sciencedirect.com/science/article/pii/S2210670721002122;  https://www.sciencedirect.com/science/article/pii/S2210670721002900

Response 2: The application of LST and the motivation of this study are illustrated. The last two years' work on surface temperature is cited and references are added. Added instructions for the quality of Landsat images selection in the data source section.

Point 3: You did use three methods to downscale, this means there is little contribution to literature. I do therefore disagree with the statement in line 67-68. Hence you must show novelty and contribution of this work to literature clearly

Response 3:We deleted the sentence in line 67-68. In the introduction, we added the innovation of the work of this paper. By comparing the downscaling results of three typical algorithms and three NDVI of GF-6 image, the effect of the newly added red-edge bands of GF-6 in the downscaling of LST was analysed and evaluated.

Point 4:You generated ground LST with instruments. What instrument was used? How long they run for? What was their reading resolution? Was there any calibration done? Nothing noted. How did you choose sitting locations? Arbitrary?

Response 4:Revised in 2.2.1. Ground observation data were obtained from three Hydrological Station (provided by Hydrographic and Water Resources Survey Bureau of Xinjiang Bortala Mongolia Autonomous Prefecture). The underlying surface around the ground station is quite homogeneous according to the field visit. The observation data measured using soil temperature probes and radiometric measurements were conducted near the ground at the station, the temporal resolutions is 1hour. The observed data at 13:00 and 14:00 (Beijing Time) were selected according to the transit time of the experimental images, and their average values were taken as the verification data.

Point 5:Table 3: how did you do normalisation? Though remote sensing community knows this well you cannot expect all readers may know this. So clarify

Response 5: We added the step of normalization into the 2.2.2. Because of the differences in radiation calibration and spectral response function with different sensors, collaborative application of multi-source sensor data cause some difficulties, and of observation geometry and atmospheric conditions impact the image. Therefore, normalization processing must be performed eliminate data discrepancies caused by these factors before the comprehensive application of multi-source remote sensing data. This process included radiometric cross-calibration [39-41], orthorectification, geometric corrections, atmospheric correction using the Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) algorithm, image cropping, and resampling. In radiometric cross-calibration, the angle information of OLI image and the slope and aspect information of ASTER GDEM product was used to establish the surface BRDF model, establish the lookup table, fit the TOA brightness of GF-6, establish the linear relationship between TOA brightness and image DN value, and fit the calibration coefficient.

Point 6: Precision metrices – this heading is wrong.

Response 6: We replaced it with “Evaluation Measures”.

Point 7: Fig. 3 – where did you get Landsat 9 LST product? Did not mention, how was it computed? Your method of LST is different from product LST? How error value in fig. 4b obtained?

Response 7: Think about the suggestion carefully, we revised the results section. Consider that Landsat-8 TIRS images from USGS have been resampled to a resolution of 30 m, GF-6 and Landsat-8 TIRS images were first scaled up to 100 meters to obtain the remote sensing index and LST under the low resolution, and to establish downscaling model. GF-6 images with original resolution were used to construct the high resolution of the remote sensing index to downscale on the low-resolution LST, and 16m LST was obtained. Landsat-8 OLI and the original TIRS images before scale-up were used to retrieve the LST, and to evaluate the downscaling results of the resampling to 30 m. So we deleted the cross validation with Landsat-8 LST product.

Point 8: Need to show info re GF satellite in supp information

Response 8: Revised. We added the information of GF satellite into the introduction

Point 9 fit lines are not well presented in all related figures, hence needing change

Response 9: Revised.

Point 10 Discussion section present but not well correlated with theory 

Response 10: Revised and reference was added. In the discussion section, the effects of the central wavelengths of GF-6 with two new red-edge bands and the spectral response of corresponding vegetation on the downscaling results of LST were discussed. 

Author Response File: Author Response.docx

Reviewer 3 Report

The reviewed manuscript focuses on the possibility to use the Gaofen-6 images for investigation of the land surface temperatures. It is rather methodological, but it also contributes to the understanding of the Ebinur Lake Watershed. The both tasks are addressed adequately. The paper is informative and novel. Its theme matches well the international research. The paper is well-written, well-organized, and well-referenced. After two rounds of its critical reading, I see only a few minor issues for amendments.

• Introduction: please, cite more works in the first paragraph.
• Subsection 2.1: I think you need to cite more literature here.
• 2: please, give caption BELOW the drawing.
• Discussion: please, indicate practical implications of your study – which tasks (regional planning, agriculture, etc.) can be solved with them?
• Why not to consider any implication to this major theme? See https://www.sciencedirect.com/science/article/pii/S0959652621015602

Author Response

Thank you for the comments. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion were mainly marked in red in the paper. The responds are as following:

Point 1: Introduction: please, cite more works in the first paragraph.

Response 1: Revised. And reference was added [2-11].

Point 2: Subsection 2.1: I think you need to cite more literature here.

Response 2: More literature were cited in subsection 2.1, and reference was added [35-38].

Point 3: please, give caption BELOW the drawing.

Response 3: Revised.

Point 4: Discussion: please, indicate practical implications of your study – which tasks (regional planning, agriculture, etc.) can be solved with them? Why not to consider any implication to this major theme? See https://www.sciencedirect.com/science/article/pii/S0959652621015602

Response 4: Revised and reference was added. Revised and reference was added. We added the problems that can be solved in the practical application of this study in the discussion. Based on analysis of the high-resolution spatial distribution of LST and its driving factors, a high-resolution geographical resilience map can be drawn, which is critical for understanding the complex interactions between human (social-economic) and natural (ecological) systems. Which is a prerequisite for environmental protection and restoration in the study area located along China's Belt and Road Initiative (BRI) economic corridors, and can be used for agricultural irrigation, regional planning, drought assessment, etc.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors did a good job and fixed most of the most questionable shortcomings of the previous version. However, there were so many changes that new defects appeared along the way. It seems that the caption for figure number 3 is "lost" somewhere. The explanation for figure number 6 is probably incomplete. I also identified a few minor bugs that were indicated in the attached PDF file.

Comments for author File: Comments.pdf

Author Response

Response to Reviewer 1 Comments

Thank you for the comments. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion were mainly marked with track changes in the paper. The responds are as following:

Point 1: The authors did a good job and fixed most of the most questionable shortcomings of the previous version. However, there were so many changes that new defects appeared along the way. It seems that the caption for figure number 3 is "lost" somewhere. The explanation for figure number 6 is probably incomplete. I also identified a few minor bugs that were indicated in the attached PDF file.


Response 1: Thank you for your affirmation. We have added the caption for Figure 3. The explanation for Figure 6 has been supplemented in the "Cross validation" and "discussion" section. All the bugs marked in the attached PDF have been revised. An English language check has been performed.

Reviewer 2 Report

Thanks, you must rectify reference section to comply with this journal's style.

Author Response

Response to Reviewer 2 Comments

Thank you for the comments. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion were mainly marked with track changes in the paper. The responds are as following:

Point 1: Thanks, you must rectify reference section to comply with this journal's style.

Response 1: We have revised the references section in accordance with the journal's style. And an English language check has been performed.