Improved Method for Estimating Construction Year of Road Bridges by Analyzing Landsat Normalized Difference Water Index 2

Round 1

Reviewer 1 Report

The study concerns the estimation of the year of construction for existing bridges based on the use of satellite Landsat data. To this purpose, the authors propose a new approach representing an evolution of past studies which is aimed to predict the construction year of bridges by using indices and algorithms of cloud removing.

The study is significant with respect to the current needs for transportation authorities which should perform screening and knowledge data collection on their assets. In this reviewer’s opinion, the methodology deserves publication. However, some modifications should be carried out, in this reviewer’s opinion, to improve the quality of the manuscript.

According to this reviewer’s opinion, sentences at Line L114 and L347 should be better clarified. Particularly, sentences explaining the technique 1 requires clarifications and improvements in language.

The state-of-the-art should be extended by referring to studies dealing with the issue of bridge safety management by using remote-sensing techniques (e.g satellites). For example, the authors should comment on the importance of research studies aimed at monitoring bridges from space, advantages and disadvantages of using satellite data and comment research trends such as extracting features from satellites (10.1109/URS.2007.371882, 10.1080/02564602.2014.892737) and monitoring via interferometry (10.1016/j.autcon.2022.104707, 10.1177/14759217211045912, 10.1007/s13349-020-00446-9).

According to this reviewer, this study includes many figures which can be redundant and make difficult for the readers understanding effectively the main outcomes of the paper. Therefore, it is suggested to add an appendix including figures showing plots and tables which can be considered secondary or redundant. In addition, necessary figures should be placed appropriately within the text as close as possible to the related sections.

 

 

The authors are invited to check the language in some parts of the text. It is suggested to use short and direct sentences which are easily understandable by readers.

Author Response

REVISIONS AND REPLIES TO THE REVIEWER 1’S COMMENTS

MDPI

Ref: remotesensing-2453936

Title: Improved Method for Estimating Construction Year of Road Bridges by Analyzing Landsat Normalized Difference Water Index 2

Authors: Bennie Hamunzala, Koji Matsumoto, and Kohei Nagai.

 

Reviewer 1

General comments:

The study concerns the estimation of the year of construction for existing bridges based on the use of satellite Landsat data. To this purpose, the authors propose a new approach representing an evolution of past studies which is aimed to predict the construction year of bridges by using indices and algorithms of cloud removing.

The study is significant with respect to the current needs for transportation authorities which should perform screening and knowledge data collection on their assets. In this reviewer’s opinion, the methodology deserves publication. However, some modifications should be carried out, in this reviewer’s opinion, to improve the quality of the manuscript.

Authors’ response: Thank you very much for your kind consideration in reviewing this paper. Your effort and time through careful reading, thoughtful suggestions, guidance, and comments towards improving our manuscript is greatly appreciated. We have revised the manuscript and have made amendments according to your recommendations.

Please refer to our responses below:

Detailed comments:

Reviewer comment: According to this reviewer’s opinion, sentences at Line L114 and L347 should be better clarified. Particularly, sentences explaining the technique 1 requires clarifications and improvements in language.

Authors’ response: Thank you for your comment and suggestions. We have clarified and made some improvements in language to Line 114-118, and Line 347-352, and the proposed changes are as follows:

 

Edited:

Line 123-126: Technique 1 states that the NDWI_2 remains the same at both the TBP and a selected optimal reference control point before the road bridge is built. After the road bridge is built, only the NDWI _2 at the TBP changes, while the NDWI_2 at a selected optimal reference control point remains the same.

Line 348-355: The current method in Figure 4b differs from the previous method in Figure 4a in that the current method (a) performs cloud masking of the data, (b) reduces cloud cover from 100% to 30%, (c) averages the data annually, (d) selects an optimal reference control point with lowest minimum variance and p-value > 0.05, and (e) uses the sequential t-test analysis of the regime shift method to interpret the estimated construction year from the NDWI_ 2. Both cloud masking and cloud cover control reduce noise in the data, and annual averaging of the data reduces seasonal variations. Homogeneity of the data is assumed when the variance is minimal and p-value > is 0.05.

 

Reviewer comment: The state-of-the-art should be extended by referring to studies dealing with the issue of bridge safety management by using remote-sensing techniques (e.g satellites). For example, the authors should comment on the importance of research studies aimed at monitoring bridges from space, advantages and disadvantages of using satellite data and comment research trends such as extracting features from satellites (10.1109/URS.2007.371882, 10.1080/02564602.2014.892737) and monitoring via interferometry (10.1016/j.autcon.2022.104707, 10.1177/14759217211045912, 10.1007/s13349-020-00446-9).

Authors’ response: Thank you for your comment and suggestions. Using the provided references, we have added some comments on the importance of research studies aimed at monitoring bridges from space, the advantages and disadvantages of using satellite data. The following are the proposed comments placed on Line 107-117:

Line 107-117: Previous studies have used remote sensing techniques with interferometric synthetic aperture radar (InSAR) and high-resolution satellites to extract features and monitor structural deformations of bridge infrastructures at the Earth’s surface. With increasing traffic loads on bridges, challenging environmental conditions such as flooding, wind loads, and other factors, and limited capital investment for bridge maintenance, repair, and rehabilitation, InSAR techniques can complement visual inspections as a source of timely information on early signs of bridge infrastructure damage. InSAR techniques can monitor infrastructure both day and night, penetrate cloud cover, produce accurate and high-resolution imagery, provide high-density measurements, satellite repetition time is relatively short, and it is a cost-effective and near real-time method. However, processing the collected InSAR data is time consuming and requires expertise. [Soergel, et al. 2007, Abraham & Sasikumar, 2015, Nettis, et al, 2023, Macchiarulo, et al. 2022, Cusson, et al. 2020, and Aswathi, et al. 2022 ].

 

Reviewer comment: According to this reviewer, this study includes many figures which can be redundant and make difficult for the readers understanding effectively the main outcomes of the paper. Therefore, it is suggested to add an appendix including figures showing plots and tables which can be considered secondary or redundant. In addition, necessary figures should be placed appropriately within the text as close as possible to the related sections.

Authors’ response: Thank you for your comment and suggestions. We have removed some Figures (6, 8, 9, 10, 13, 14, and 17) and Tables (4, and 6 ) that we considered to be redundant. We also moved some Figures to an appendix and placed the Figures and Tables as close as possible to the related sections. Please refer to the edited version.

 

Comments on the Quality of English Language

Reviewer comment: The authors are invited to check the language in some parts of the text. It is suggested to use short and direct sentences which are easily understandable by readers.

Authors’ response: Thank you for your comment and suggestions. We edited the whole document and used some direct sentences where we thought it was appropriate. Please refer to the edited version.

References

Soergel, U.; Thiele, A.; Gross, H.; Thoennessen, U. Extraction of bridge features from high-resolution InSAR data and optical images. Urban Remote Sensing Joint Event 2007, 1–6.

Abraham, L.; Sasikumar, M. Analysis of satellite images for the extraction of structural features. IETE Technical Review 2014, 31(2), 118–127.

Nettis, A.; Massimi, V.; Nutricato, R.; Nitti, D. O.; Samarelli, S.; Uva, G. Satellite-based interferometry for monitoring structural deformations of bridge portfolios. Automation in Construction 2023, 147, 104707.

Macchiarulo, V.; Milillo, P.; Blenkinsopp, C.; Giardina, G. Monitoring deformations of infrastructure networks: A fully automated GIS integration and analysis of InSAR time-series. Structural Health Monitoring 2022, 21(4), 1849–1878.

Cusson, D.; Rossi, C.; Ozkan, I. F. Early warning system for the detection of unexpected bridge displacements from radar satellite data. Journal of Civil Structural Health Monitoring 2021, 11, 189–204.

Aswathi, J.; Binojkumar, R. B.; Oommen, T.; Bouali, E. H.; Sajinkumar, K. S. InSAR as a tool for monitoring hydropower projects: A review. Energy Geoscience 2022.

 

 

 

Reviewer 2 Report

The research findings indicate that both methodologies were successful in estimating the construction year of road bridges, albeit with varying degrees of accuracy.

 

Technique 1, which employs a Target Bridge Point (TBP) and selected optimum reference control points, yielded accurate results with a cut-off length of ≤9, an absolute error of ≤5, and statistical significance with p-values <0.05.

 

 

 

In contrast, Technique 2, which relies solely on the TBP, produced inaccurate results for three road bridges, a statistically insignificant result for one bridge, and no results for five bridges. The primary causes of these inaccurate and absent results were seasonal variations and insufficient reflectance at the TBP.

 

The correlation analysis revealed an increase in the absolute error between the actual and estimated construction year as the cut-off length increased. Consequently, to ensure accuracy, the upper limit of the cut-off length was set to ≤12.

 

The research also discovered that the accuracy of results is influenced by the increase in the overall length (in the range of 30 m to 100 m) and the presence of forested and cropland areas. All five bridges that yielded no results in Technique 2 had an overall length of less than 30 m. Similarly, out of the thirteen bridges that yielded no results in Technique 1, eight had an overall length of less than 30 m.

 

Here are some questions about these results:

 

What criteria were used to select the TBPs and the optimum reference control points for Technique 1? How was their suitability determined?

 

The authors are invited to identify specific factors that led to the inaccurate results for the three bridges in Technique 2? Did these bridges share any common characteristics that might explain the inaccuracies?

 

The authors need to provide a mathematical or statistical explanation for how the cut-off length impacts the accuracy of the estimated construction year? Why was a cut-off length of ≤12 chosen as the upper limit?

 

The results suggest that the overall length of the bridge and the surrounding environment (forested and cropland areas) contribute to the accuracy of the results. Can you elaborate on why these factors are significant?

 

How might the methodologies be adjusted to account for the identified challenges of seasonal variations and inadequate reflectance at the TBP?

 

How might these methodologies be applied or adapted for use in other geographic areas or for other types of infrastructure?

 

The English quality of the paper appears to be good, with clear and concise language used throughout. The authors have used appropriate scientific terminology and have structured the paper in a logical manner, with sections for the methodology, results, and discussion. The figures and tables are well-labeled and contribute to the overall understanding of the research.

Author Response

REVISIONS AND REPLIES TO THE REVIEWER 2’S COMMENTS

 

MDPI

Ref: remotesensing-2453936

Title: Improved Method for Estimating Construction Year of Road Bridges by Analyzing Landsat Normalized Difference Water Index 2

Authors: Bennie Hamunzala, Koji Matsumoto, and Kohei Nagai.

 

Reviewer 2

General comments:

The research findings indicate that both methodologies were successful in estimating the construction year of road bridges, albeit with varying degrees of accuracy.

Technique 1, which employs a Target Bridge Point (TBP) and selected optimum reference control points, yielded accurate results with a cut-off length of ≤9, an absolute error of ≤5, and statistical significance with p-values <0.05.

In contrast, Technique 2, which relies solely on the TBP, produced inaccurate results for three road bridges, a statistically insignificant result for one bridge, and no results for five bridges. The primary causes of these inaccurate and absent results were seasonal variations and insufficient reflectance at the TBP.

The correlation analysis revealed an increase in the absolute error between the actual and estimated construction year as the cut-off length increased. Consequently, to ensure accuracy, the upper limit of the cut-off length was set to ≤12.

The research also discovered that the accuracy of results is influenced by the increase in the overall length (in the range of 30 m to 100 m) and the presence of forested and cropland areas. All five bridges that yielded no results in Technique 2 had an overall length of less than 30 m. Similarly, out of the thirteen bridges that yielded no results in Technique 1, eight had an overall length of less than 30 m.

 

 

Authors’ response: Thank you very much for your kind consideration in reviewing this paper. Your effort and time through careful reading, thoughtful suggestions, guidance, and comments towards improving our manuscript is greatly appreciated. We have revised the manuscript and have made amendments according to your recommendations.

Please refer to our responses below:

Detailed comments:

Reviewer comment: What criteria were used to select the TBPs and the optimum reference control points for Technique 1? How was their suitability determined?

Author response: Thank you for your questions. We have attempted to respond to your questions as follows:

Line 272-323:  The flowchart for the current method is shown in Figure 4b. The current method consists of both Technique 1 and Technique 2. Technique 1 uses both the TBP and a selected optimal reference control point, while Technique 2 only uses the TBP, to determine the estimated construction year of the road bridge. Japan has about 700,000 road bridges. From about 700,000 road bridges, 330 road bridges in Nago City were filtered out. Nago City is located in the northern part of Okinawa Prefecture. From 330 road bridges, 44 road bridges were filtered out based on the bridge’s overall length between 0 m and 100 m and construction year between 1990 and 2006, as shown in Table 4.

In Step 1 and Step 2, we used Google Earth Engine (GEE) and global positioning system (GPS) coordinates, from the Japan bridge database [Japan bridge database], to locate the midspan of the bridge’s overall length for all 44 road bridges, which we called the target bridge point (TBP). After we located the TBP in Google Earth Engine (GEE), the new GPS coordinates at the midspan were collected and entered in GEE. A perpendicular line to the bridge’s overall length was drawn 60 m from the TBP and 18 reference control points were established, 9 points on each side of the TBP. Each reference control point and TBP occupied an independent and unique image pixel with a spatial resolution of 30 m, as shown in Figure 7a and Figure 7b, and Figure A1a and Figure A1b, for Kuroyama road bridge, and Kamiyama road bridge, respectively. The 60 m distance was chosen because it had less interference with the TBP image pixel. The GPS coordinates for the 18 reference control points were collected and entered into GEE. GEE is a cloud computing program for geospatial data analysis and visualization [Google Earth Engine].

In Step 3, we analyzed the cloud unmasked data at 100% cloud cover in GEE at the TBP and 18 reference control points, as shown in Figure 7c, and Figure A1c, for Kuroyama road bridge, and Kamiyama road bridge, respectively. Landsat 5 Thematic Mapper (TM) was used for the analysis. Cloud unmasked data are data points whose pixels are contaminated with clouds and cloud shadows and therefore obscure the visibility of the surface cover due to the noise in the data.

In Step 4, we used GEE to cloud mask the data at 100% cloud cover, at the TBP and 18 reference control points, as shown in Figure 7d and Figure A1d for Kuroyama road bridge, and Kamiyama road bridge, respectively. Landsat 5 Thematic Mapper (TM) was used for the analysis. The cloud cover was reduced from 100% to 30%, and the data were averaged annually, as shown in Figure 8a, and Figure A3a, for Kuroyama road bridge, and Kamiyama road bridge, respectively. Both cloud masking and cloud cover control reduce noise in the data, and annual averaging of the data reduces seasonal variations.

In Step 5, we calculated the p-values and variances at the TBP and 18 reference control points. This step was performed to check the normality and variation of the data points at the TBP and the reference control points. Homogeneity of the data was assumed when the variance was minimal and the p-value > 0.05.

In Steps 6 and Step 7, we selected an optimal reference control point from the 18 reference control points. The reference control point with the lowest minimum variance and p-value > 0.05 was considered an optimal reference control point. If two or more reference points had the same variance values and p-value> 0.05, the reference control point farthest from the TBP was selected as the optimal point because it was considered to be least affected by the interference from the TBP. The TBP and a selected optimal reference control point were plotted together, as shown in Figure 8a, and Figure A3a, for Kuroyama road bridge, and Kamiyama road bridge, respectively.

In Step 8a, Technique 1, which uses both the TBP and a selected optimal reference control point, was used to determine the estimated construction year. The sequential t-test analysis of regime shift (STARS) method [Rodionov, et al. 2005] was used to the interpret the estimated construction year from NDWI_ 2 for squared difference between TBP and a selected optimal reference control point, as shown in Figure 8b and Figure 8d, and Figure A3b and Figure A3d, for Kuroyama road bridge, and Kamiyama road bridge, respectively.

 

Reviewer comment: The authors are invited to identify specific factors that led to the inaccurate results for the three bridges in Technique 2? Did these bridges share any common characteristics that might explain the inaccuracies?

Authors’ response: Thank you for your questions. We have attempted to respond to your questions as follows:

Line 537-541:  All three road bridges with inaccurate results had seasonal variations that could have contributed to the inaccurate results. Of the three road bridges, Fukumi had excessive seasonal variations. The overall length of both Maehira road bridge and the Kayang Mata road bridge was less than 30 m and may have contributed to inaccurate results in the STARS method due to inadequate reflectance.

 

Reviewer comment: The authors need to provide a mathematical or statistical explanation for how the cut-off length impacts the accuracy of the estimated construction year? Why was a cut-off length of ≤12 chosen as the upper limit?

Authors’ response: Thank you for your questions. We have attempted to respond to your questions as follows:

Line 486-496:  The increase in regime shift index RSI had an insignificant effect on the absolute error between the actual and estimated construction year, as shown in Figure 13a, and Figure 13b. However, as the cutoff length l increased, the absolute error between the actual and estimated construction year increased, as shown in Figure 13c and Figure 13d. The degree of freedom also increases as the cutoff length l increases, leading to a decrease in the statistically significant difference between the means of two successive regimes [Rodionov, et al. 2005, Rodionov, et al. 2004], as shown in Equations (1) – (3). To increase the accuracy of the estimated construction year, we may consider setting the upper limit of the cutoff length l to l ≤ 12 from Figure 13d. Based on the clustered results with minimum absolute errors in Figure 13d, l ≤ 12 was selected as the working limit from the linear regression analysis between the cutoff length l and the absolute error between the actual and estimated construction year, for Technique 2.

 

Line 574-575:  The cutoff length l ≤ 12 is a working limit, therefore, in the future, there is a need to develop a robust reliability method for these proposed Techniques.

 Line 614-616:  The cutoff length l ≤ 12 is a working limit, therefore, in the future, there is a need to develop a robust reliability method for these proposed Techniques.

 

Reviewer comment: The results suggest that the overall length of the bridge and the surrounding environment (forested and cropland areas) contribute to the accuracy of the results. Can you elaborate on why these factors are significant?

Authors’ response: Thank you for your questions. We have attempted to respond to your questions as follows:

 

Edited:

Line 467-485: For Technique 2, R² = 0.41 was higher for medium road bridges than R² = 0.20 for short road bridges, and R² = 0.42 was higher for road bridges in forested and cropland areas than R² = 0.05 for road bridges in built-up areas, as shown in Table 5. The results suggest that a greater overall length of road bridges and road bridges in forested areas contributes to accuracy. One of the possible explanations is that a road bridge with a greater overall length and a reasonable road width has a large surface area that results in a higher reflectance value, provided that the reflectance is minimally attenuated by scattering and absorption effects and the image pixel is sufficiently covered by the road bridge surface. No analysis of the relationship between road width and accuracy was performed in this study. The normalized difference water index 2 (NDWI_2) is a function of near-infrared band (NIR) and shortwave infrared band 2 (SWIR2). Green healthy vegetation has a higher reflectance value in the NIR than in the SWIR2, and green vegetation also has a higher reflectance value than asphalt or concrete surfaces in the NIR [Herold, et al. 2003]. Therefore, the reflectance is expected to be higher before the bridge is built, assuming that the original surface cover was green healthy vegetation. After the road bridge is built, the reflectance decreases significantly, which in turn increases the mean difference between the current and new regimes in the sequential t-test analysis of the regime shift method assuming that the reflectance is minimally attenuated due to scattering and absorption effects.

 

Reviewer comment: How might the methodologies be adjusted to account for the identified challenges of seasonal variations and inadequate reflectance at the TBP?

Authors’ response: Thank you for your question. We have attempted to respond to your question as follows:

Line 571-574: In future research, some correction factors could be determined using statistics or mathematical formulas to account for seasonal variations and inadequate reflectance at the TBP, as well as using all image pixels across the whole span of the TBP to increase reflectance.

 

Reviewer comment: How might these methodologies be applied or adapted for use in other geographic areas or for other types of infrastructure?

Authors’ response: Thank you for your question. We have attempted to respond to your question as follows:

Line 630-634: The proposed Techniques can be applied to detect any significant mean difference on the Earth’s surface, whether natural or man-made. The region must be covered by Landsat or related satellites, the features on the Earth’s surface must not be obscured by natural or human cover for extended periods, and the data from Landsat or related satellites must not be severely attenuated by atmospheric absorption and scattering effects.

Comments on the Quality of English Language

Reviewer comment: The English quality of the paper appears to be good, with clear and concise language used throughout. The authors have used appropriate scientific terminology and have structured the paper in a logical manner, with sections for the methodology, results, and discussion. The figures and tables are well-labeled and contribute to the overall understanding of the research.

Authors’ response: Thank you for your comments.

 

 

References

National road facility inspection database for Japan. Available online: https://road-structures-map.mlit.go.jp/FacilityList.aspx (accessed on 21 December 2022)

Google Earth Engine. Available online: https://www.google.com/earth/education/tools/google-earth-engine/ (accessed on 19 December 2022).

Rodionov, S.N.; Overland, J.E. Application of a sequential regime shift detection method to the Bering Sea ecosystem. ICES Journal of Marine Science 2005, 62, 328–332 

A Herold, M.; Gardner, M. E.; Noronha, V.; Roberts, D. A. Spectrometry and hyperspectral remote sensing of urban road infrastructure. Online Journal of Space Communication 2003, 2(3), 9.

Rodionov, S.N. A sequential method for detecting regime shifts in the mean and variance. Large-scale disturbances (regime shifts) and recovery in aquatic ecosystems: challenges for management toward sustainability 2005, 68–72.

Rodionov, S.N. A sequential algorithm for testing climate regime shifts. Geophysical Research Letters 2004, 31.

 

 

 

 

 

 

  

 

Reviewer 3 Report

Based on the relevant academic literature and theoretical research, this paper analyzes the existing methods, improves the NDWI_2 method, and puts forward two techniques for estimating the construction year of highway Bridges, which provide the basis for the maintenance and improvement of highway Bridges. The whole idea of the thesis is clear, the method is desirable, the content is detailed and solid, and the topic selection has certain theoretical value and practical significance. It is shown from Table 7 and Figure 23 that coefficient of determination R² are too small.

Author Response

REVISIONS AND REPLIES TO THE REVIEWER 3’S COMMENTS

 

MDPI

Ref: remotesensing-2453936

Title: Improved Method for Estimating Construction Year of Road Bridges by Analyzing Landsat Normalized Difference Water Index 2

Authors: Bennie Hamunzala, Koji Matsumoto, and Kohei Nagai.

 

Reviewer 3

General comments:

Based on the relevant academic literature and theoretical research, this paper analyzes the existing methods, improves the NDWI_2 method, and puts forward two techniques for estimating the construction year of highway Bridges, which provide the basis for the maintenance and improvement of highway Bridges. The whole idea of the thesis is clear, the method is desirable, the content is detailed and solid, and the topic selection has certain theoretical value and practical significance. It is shown from Table 7 and Figure 23 that coefficient of determination R² are too small.

Authors’ response: Thank you very much for your kind consideration in reviewing this paper. Your effort and time through careful reading, thoughtful suggestions, guidance, and comments towards improving our manuscript is greatly appreciated.  We have attempted to respond to your comments as follows:

 Line 455-463: Table 5 shows the summary of the results of the correlation analysis. Correlation analysis yielded R² = 0.24 and R² = 0.33 and an average deviation of S = 5.81 years and S = 4.08 years for Technique 1 and Technique 2, respectively, as shown in Figure 12a and Figure 12b, respectively. The results suggest that Technique 2 is more accurate and provides a better estimate than Technique 1, but R² values are low. One of the possible explanations that may have contributed to low R² values, as indicated by the gap between the actual and estimated year of construction in Figure 12, is that the planned year of construction recorded in the database may not correspond to the actual year of construction due to construction delays [Nagai, et al. 2023].

 Line 605-611:  Correlation analysis of all 44 road bridges yielded R² = 0.24 and R² = 0.33 and an average deviation of S = 5.81 years and S = 4.08 years for Technique 1 and Technique 2, respectively. The results suggest that Technique 2 is more accurate and is a better estimate than Technique 1, but R² values are low. One of the possible explanations that may have contributed to low R² values, as indicated by the gap between the actual and estimated year of construction, is that the planned year of construction recorded in the database may not correspond to the actual year of construction due to construction delays.

 

References

Sovisoth, E.; Kuntal, V.S.; Misra, P.; Takeuchi, W.; Nagai, K. Estimation of Year of Construction of Bridges in Cambodia by 623 Analyzing the Landsat Normalized Difference Water Index. MDPI-Infrastructures 2023, 8(4), 77.

Reviewer 4 Report

This article presents interesting results of a research about a very important topic, which has not been sufficiently studied in the worldwide literature. The article is well written and organized, the approach is innovative and its quality is high. Moreover, in general, figures and tables are clear and bibliography is adequate. As far as the knowledge of the Reviewer, the results are published here for the first time. Therefore, I recommend this article to be published after a minor revision:

 

* There are several typos that must be corrected, some of them are shown below. Please, recheck the entire article.

- Blank spaces between a value and the sign of “per cent” (%) are included and this is not correct; for example, page 3 line 141 (4.3 %).

- There is a typo in captions of Figures 21-23: “the regression which represents” instead of “the regression, which represents”.

- The authors should use “cannot” instead of “can not” in the entire article.

- The authors sometimes use “normalized” and others “normalised”. It is necessary to use only American English or British English, but not mix them.

 

* In Figure 2, the legend for “Y” axis is correct just for Figure 2a (Population count), whereas for Figure 2b to Figure 2e is not correct using just “Count”. On the other hand, for the legend of “X” axis, in Figure 2a and Figure 2b the word “Cities” should be included below the different cities.

 

* In Figure 2f, the dark green color should be changed for another color tone to observe better the words inside that part.

 

* In the Abstract and Introduction, the authors mention clearly that they use 2 techniques (Technique 1 and Technique 2) for the purpose of the article and they are compared each other; however, after that, sometimes it is confusing which one is Technique 1 and Technique 2. For example, in Section 2, a figure (Figure 4) was included with the caption “Method flowchart for the previous and current work”, but it is confusing if “current work” correspond with Technique 1, Technique 2 or both, that should be clearly mentioned because it is very important to include schematic diagrams of both techniques so that the readers can know at a glance both techniques. Moreover, the authors must indicate clearly, which results correspond with Technique 1 and Technique 2 during the entire article (including those presented in Figures and Tables).

 

* Captions of (c) and (d) for Figure 5 to Figure 10, must be more detailed and include the word “Year” for legends of “X” axes.

 

* In Table 3, the authors must indicate the “Perpendicular distance from TBP” corresponding with Technique 1 and Technique 2 and use the same significant digits.

 

* In plots of Figure 11 to Figure 14, and Figure 16 to Figure 20, the word “Year” for legends of “X” axes must be included.

  

* In captions of Figure 11 to Figure 14, the term “constr. year” was mentioned, but in the plots that term is called “constr. date”.

 

* In Section 4 (Discussion), a comparative table must be included, where the best technique proposed in this article is compared against other works (with references included) using similar and different techniques for the same purpose and highlighting the advantages of the proposed methodology.

Minor editing of English language required.

Author Response

REVISIONS AND REPLIES TO THE REVIEWER 4’S COMMENTS

 

MDPI

Ref: remotesensing-2453936

Title: Improved Method for Estimating Construction Year of Road Bridges by Analyzing Landsat Normalized Difference Water Index 2

Authors: Bennie Hamunzala, Koji Matsumoto, and Kohei Nagai.

 

Reviewer 4

General comments:

This article presents interesting results of a research about a very important topic, which has not been sufficiently studied in the worldwide literature. The article is well written and organized, the approach is innovative and its quality is high. Moreover, in general, figures and tables are clear and bibliography is adequate. As far as the knowledge of the Reviewer, the results are published here for the first time. Therefore, I recommend this article to be published after a minor revision:

Authors’ response: Thank you very much for your kind consideration in reviewing this paper. Your effort and time through careful reading, thoughtful suggestions, guidance, and comments towards improving our manuscript is greatly appreciated. We have revised the manuscript and have made amendments according to your recommendations.

Please refer to our responses below:

Detailed comments:

Reviewer comment: * There are several typos that must be corrected, some of them are shown below. Please, recheck the entire article.

- Blank spaces between a value and the sign of “per cent” (%) are included and this is not correct; for example, page 3 line 141 (4.3 %).

- There is a typo in captions of Figures 21-23: “the regression which represents” instead of “the regression, which represents”.

- The authors should use “cannot” instead of “can not” in the entire article.

- The authors sometimes use “normalized” and others “normalised”. It is necessary to use only American English or British English, but not mix them.

Authors’ response: Thank you for your observations and the following are revisions:

• We have corrected the spaces between the figure and the percent sign.
• We have corrected the typo in the captions of Figure 21 – 23 (currently Figure 10 – 12): “the regression which represents” instead of “the regression, which represents”.
• We have replaced the can not with cannot.
• We have replaced normalised with normalized.

 

Reviewer comment: * In Figure 2, the legend for “Y” axis is correct just for Figure 2a (Population count), whereas for Figure 2b to Figure 2e is not correct using just “Count”. On the other hand, for the legend of “X” axis, in Figure 2a and Figure 2b the word “Cities” should be included below the different cities. *And in Figure 2f, the dark green color should be changed for another color tone to observe better the words inside that part.

Authors’ response: Thank you for your observations and the following are the revisions:

• We have revised Figure 2a by adding the word “City” after all the 9 cities on the y-axis legend.
• We have revised Figure 2b by adding the word “City” after all the 9 cities on the x-axis legend and replacing “Count” with “Bridge count and surface area (km²)” on the x-axis legend.
• We have revised Figure 2c by replacing “Count” with “Bridge count” on the y-axis legend.
• We have revised Figure 2d by replacing “Count” with “Bridge count” on the y-axis legend.
• We have revised Figure 2e by replacing “Count” with “Bridge count” on the y-axis legend.
• We have revised Figure 2f by changing the dark green color to a lighter one and also included the white background for the text in the pie chart.

 

Reviewer comment: * In the Abstract and Introduction, the authors mention clearly that they use 2 techniques (Technique 1 and Technique 2) for the purpose of the article and they are compared each other; however, after that, sometimes it is confusing which one is Technique 1 and Technique 2. For example, in Section 2, a figure (Figure 4) was included with the caption “Method flowchart for the previous and current work”, but it is confusing if “current work” correspond with Technique 1, Technique 2 or both, that should be clearly mentioned because it is very important to include schematic diagrams of both techniques so that the readers can know at a glance both techniques. Moreover, the authors must indicate clearly, which results correspond with Technique 1 and Technique 2 during the entire article (including those presented in Figures and Tables).

Authors’ response: Thank you for your observations and the following are the revisions:

• We have revised Figure 4b by including Technique 1 and Technique 2 within the same flowchart at the very end, Step 8a for Technique 1 and Step 8b for Technique 2.
• We revised the sub-caption for Figure 4b to clarify that the current method is consists of both Technique 1 and Technique 2.
• We changed the sub-caption for Figure 4b from “Method flowchart for the current work” to “Method flowchart for the current work consists of both Technique 1 and Technique 2”.
• We revised the main caption for Figure 4 by adding some details about Technique 1 and Technique 2.

 

Reviewer comment: * Captions of (c) and (d) for Figure 5 to Figure 10, must be more detailed and include the word “Year” for legends of “X” axes.

Authors’ response: Thank you for your observations and the following are the revisions:

• We added more details to the sub-captions of (c) and (d) for Figure 5 to 10 (currently Figures 7, A1, and A3) and included the “Year” for the x-axis legend.
• We revised the main captions for Figure 5 to 10 (currently Figures 7, A1, and A3) by adding some details about Technique 1 and Technique 2.
• We also decided to remove some Figures (6, 8, 9, 10, 13, 14, and 17) that we considered redundant, and moved the other Figures to the appendix.

 

Reviewer comment: * In Table 3, the authors must indicate the “Perpendicular distance from TBP” corresponding with Technique 1 and Technique 2 and use the same significant digits.

Authors’ response: Thank you for your observations. We revised the whole Table 3 and included additional column, so that Technique 1 and Technique 2 are clearly shown. We also included additional notes just below Table 3.

 

Reviewer comment:

* In plots of Figure 11 to Figure 14, and Figure 16 to Figure 20, the word “Year” for legends of “X” axes must be included.

* In captions of Figure 11 to Figure 14, the term “constr. year” was mentioned, but in the plots that term is called “constr. date”.

 Authors’ response: Thank you for your observations.

• We revised and included the word “Year” for legends of the x-axis legend and the replaced “constr. date” with “constr. year” for Figure 11 to Figure 14 (currently Figure 6 and A5), and Figure 16 to Figure 20 (currently Figure 8, 9, and A2). For Figure 21 to Figure 23 (currently Figure 10 – 12), we included the word “Year” for legends of x-axis legend and y-axis legends. We are also added the Technique 1 and Technique 2 for the mentioned Figures including Figure 24 (currently Figure 13).
• We also decided to remove some Figures (6, 8, 9, 10, 13, 14, and 17) that we considered redundant, rearranged some them, and moved the other Figures to the appendix.

 

Reviewer comment:

* In Section 4 (Discussion), a comparative table must be included, where the best technique proposed in this article is compared against other works (with references included) using similar and different techniques for the same purpose and highlighting the advantages of the proposed methodology.

 Authors’ response: Thank you for your observations. We included a comparative table (Table 6) and highlighted the advantages of the proposed methodology.

 

Comments on the Quality of English Language

Reviewer comment:

Minor editing of English language required.

Authors’ response: Thank you for your observations. We edited the English in the whole document. 

Round 2

Reviewer 2 Report

The manuscript is in good order, with all necessary corrections and enhancements duly integrated. It is now ready for publication.