Continuous Monitoring of Urban Land Cover Change Trajectories with Landsat Time Series and LandTrendr-Google Earth Engine Cloud Computing

Round 1

Reviewer 1 Report

Continuous urban land cover change trajectories were tested and developed with two dates of land use maps at the end dates of the period and time series Landsat. The manuscript is well-structure and written, but a critical part is missing for describing the algorithm of Landtrendr. The algorithm and the 8 associated parameters for constructing fitted model need to be explained, description in in section 1.1 and figure 1 is not enough because Landtrendr is the main algorithm you implemented.  The specific concerns are list below.

1. Page 2, line 63, a comma is missed between “manage” and “and”.
2. Page 2, line 87, there should be citation as example for urban development study with time series EO data.
3. Page 3, line 127 to 186, instead of having a subsection in introduction, a new section independent from introduction is better in terms of paper structure.
4. Page 3, line 153, use “of” instead of “to” after findings.
5. Page 3, line 170 to 172, this sentence needs a citation.
6. Page 5, line 204, “road” should be adjective, so remove “s”.
7. Page 6, Figure 2, the background Landsat image should change to other color composite for discrimination among built, water, and veg. Color infrared is a better composite for such purpose.
8. Page 6, line 217 and 222, the information is conflict. TOA is mentioned in line 217, while surface reflectance in line 222. TOA and surface reflectance are different reflectance measurements. Why use TOA for the 1987 image while surface reflectance for the 2019 image? Surface reflectance product should be available for both dates in Google earth engine.
9. Page 6, line 6, 225 to 226, what is the source of the DTM, is that from SRTM, or other local source? The spec of the DTM (e.g. spatial resolution) need to be manifested.
10. Page 7, line 237 to 238, you do not need to conduct image registration if both images are from level-2 surface reflectance.
11. Page 7, line 243, RMSE was suggested smaller than 0.5 pixel, and a reason must be provided why RMSE is over 0.5 pixel.
12. Page 7, line 248, how did you determine the 11 by 11 kernel size is appropriate for such purpose?
13. Page 8, line 301, the link for TCG and TCW calculation are for Landsat 7. Are you using the same coefficients for calculating TCG and TCW for all Landsat 5, 7, and 8 imagery? I am worried whether you should use the same coefficients for Landsat 8 imagery since their blue band specification of Landsat 8 from the specification for Landsat 5 and 7. If so, you should have a citation for justify the coefficient adoption.
14. Page 9, figure 4, some acronyms need to be explained in the caption. For example, DUR, YOD, MAG.
15. Page 10, line 337, “to” is missing behind ‘PCA”. The sentence is too complicated and containing too much information. Breaking the sentence to two shorten sentences would be better clarify the idea.
16. Page 11, figure 6, YOD needs to be explained in the caption. What does the bracket indicate for?
17. Page 12, line 388, removing change samples is problematic for a change detection-based study. The change samples are representative for the land change products, and some studies even tried to allocate more samples in the change areas. time saving is not a solid reason for removing change samples during accuracy assessment. I suggest incorporating more change sample for strengthening the findings of the study.
18. Page 12, equation 5, why use “?” for user’s accuracy? Isn’t “U” a more intuitive notation for representing user’s accuracy?
19. Page 13, line 396 to 398, this paragraph is redundant. Remove it.
20. Page 13, Table 1, why the 1987 map are more accurate than the 2019 map? I really concern the results of Table 1 because maps derived at recent year tend to have high accuracy than those derived from early period in past studies. I wonder whether the error or mis-estimation comes from the sampling strategy and the change sample removal. The difference in accuracy needs to be explained.
21. Page 13, line 400 -401 and line 407 -408, these sentences should be merged into one paragraph with appropriate connection since the lengths are short when they are separated.
22. Page 14, Table 2, how did you determine the 8 values are the optimal values? What is the judge criterion?
23. Page 15, line 422 to 423, is this comes from visually interpret on time series plots of NDVI and TCG for the change areas? Or, the high change magnitude since 2008 was ensured based on Google earth VHR imagery?
24. Page 15, line 422 to 427, provide multiple time series plot for representing different land cover change in a single figure will give audience better concept than only describing the time series plots in text.
25. Page 15, line 430, a city is passively be observed for land cover conversion, not in active way. Change the verb accordingly.
26. Page 15, line 431, “extensive” is more appropriate than “intensive” in terms of urban land growth.
27. Page 16, line 462, typo in “2014” for “2004”? The largest urban increment is confirmed in figure 9 and Table 4.
28. Page 16, line 465, typo in “20013” for “2013”?
29. Page 20, line 484 to 487, this section is redundant, remove it.

Author Response

We truly appreciate your constructive comments and efforts toward improving our manuscript. In the attached file we explain how we took care of your comments and suggestions.

Sincerely

Author Response File: Author Response.pdf

Reviewer 2 Report

Landtrendr is widely used for monitoring the trends of land cover disturbance and recovery, and this paper employed this tool to reconstruct progressive dense land cover maps in a certain period, taking Kigali, Rwanda as a case. Compared with previous studies, the main innovation of this paper is using a combination of indices instead of single indices. The topic is quite interesting and valuable. There are some comments and concerns as follows:

(1) The paper mentioned several times that the proposed method is a cost-effective method. How to prove the method “cost-effective” relative to previous studies, there should be some data to support the judgement.

(2)What is the role of the combined index? Only for the classification of land cover types? In my opinion, the combined index should have some certain meanings and whose variation could reflect the process of suffering disturbance and recovery (for each pixel). But in this paper, the authors did not discuss the change of the index, instead, they use this index to guarantee the precise of the classification. For example, in the conclusion part, the authors mainly spend their words to describe the increase or decrease of the urban spaces, while none for the index.

(3) Previous studies demonstrated that Overall Accuracy (OA) may not be an adequate measure in evaluating the accuracy of a classification when the dataset is imbalanced, e.g. when a certain land cover type account for the vast majority of the area. More evaluation indicators should be added to support your views.

(4) The selection of the six indices is too subjective. There is a certain correlation between some indices, that is why you can obtain a good PCA result. I wonder if removing one or two of them will affect the accuracy of classification. Again, It would be better if the synthesized index were meaningful.

(5)Some minor mistakes:

• In Figure 9. “Year od detection” should be “Year of detection”
• line 479, “215” should be “2015”.

Author Response

We truly appreciate your constructive comments and efforts. You will find in the attached file point-by-point responses to all the good questions and inquiries for improving our manuscript

Author Response File: Author Response.pdf

Reviewer 3 Report

Authors have conducted interesting work from the scientific viewpoint. The approach is well founded and the results are interesting. However, I found few issues to be addressed before it gets published.

1. L13-16: “However… trends”. This sentence is very confusing. Please re-write it.
2. L26-27: What are these terms? You did not tell anything about them in the introduction.
3. L118-119: Not a scientific way of expressing thanks. You may cite the organization or acknowledge it.
4. L127: Why was it necessary to highlight in a subsection if there is no 1.2 section?
5. L173-174: What “others”? This statement is not consistent with the abstract. Please explain each term in plain text?
6. L181-186: Authors have highlighted many things and many are not relevant to the present objective. Please remove the text and citations which are not appropriate to the objective.
7. Objective is very poorly defined. It is advised to highlight more explicitly what authors would like to do.
8. L243: What is GLCM? Please explain it in more detail.
9. L279: What are Spectral-temporal metrics? How did you do that? Which bands did you consider?
10. L290-296: Similar information should be reflected in plain text in the Introduction.
11. L358: Eq 3. Do you have a reference for this equation?
12. L393: Eq. 5: Is this question mark “?” symbol correct?
13. L395: Results section must be explained in more detail. How are your findings of using combined indices effective in monitoring the complex urban land cover dynamics? It is very important to highlight how your proposed methodology of using combined indices performs differently from those existing methods.
14. Table 1: What are these covers? Please define these land cover types somewhere in the methods section. What about other covers e.g. grassland, shrubs, bare land?
15. Table 1: Why the values under the User’s accuracy for Urban and Forest in 1987 have low values? Do you have any explanation? Perhaps a limitation of the proposed method.
16. Table 4: Reference numbers may confuse the readers. Change reference numbers to land cover types. Authors can use initials for each land cover. I am also wondering why there is no change between 3 and 4 and negligible change between 4 and 5? Similar comments to all other years also.
17. L502-504: It is advisable to discuss, how the combined indices outperform the single indices?
18. Fig. 2, 5, 7, 8. Axis values look hazy. Make it in decimal °S, °E, e.g. 1°S 1.5° 2°S.
19. Many references. Must be reduced.

Author Response

We truly appreciate your constructive comments and efforts: you will find in the attached file point-by-point responses to all the good questions and inquiries for improving our manuscript

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

This version of manuscript is much improved. One minor issue is associated with the equation 5. I still saw "?" for representing user's accuracy from my side. Please double-check the issue is fixed.

Author Response

Thank you very much for the  constructive review. The "?" was resulting to PDF conversion which went wrong. It was now corrected. We suggested to submit a separate file with the proper equation in case the  conversion to PDF is still affecting the equation write-up

Reviewer 2 Report

 I suggest accepting the paper in the current form. No more comments and concerns.

Author Response

Thank you very much for constructive comments that were helpful in improving our manuscript

Reviewer 3 Report

Authors have considered all my comments in the revised manuscript. There have also been substantial improvements found in the revised manuscript. It may be possible to accept the article for publication after following few minor corrections.

1. Table 1: 1^st column i.e. the “No” column is not necessary. It may be deleted.
2. Line 428- Equation 5: The wrong symbol “?” should be replaced with correct symbol.
3. Line 429-430: There is a Chinese character. It should be checked and removed.
4. Figure 7: Please check the South “S” and East “E” marks in the right sight figure.

Author Response

Thank you very much for constructive comments that were helpful in improving our manuscript. All suggested revisions were addressed in the manuscript.

1.The “No” column was removed as can be seen in the Table 1

2.The wrong symbol “?” in the equation 5 was  replaced with correct symbol. this can be read in Equation 5. The error was caused by the conversion of the word document to PDF. In case the same error persists, we suggest to submit a separate file with proper written Equation

3.There is a Chinese character. It was checked and removed. new correct writing can be seen in Equation 5. The error was caused by the conversion of the word document to PDF. In case the same error persists, we suggest to submit a separate file with proper written Equation

4. The South “S” and East “E” marks in the right sight figure was correctly addressed. Check the Figure 7