Urbanization and Its Impacts on Land Surface Temperature in Colombo Metropolitan Area, Sri Lanka, from 1988 to 2016
Round 1
Reviewer 1 Report
Overall comments
This manuscript by Fonseka et al. explored the urban expansion from 1988 to 2016 in Colombo Metropolitan Area, Sri Lanka using multi-temporal Landsat images. Two methods have been used to evaluate the impact of rapid urbanization on the Land Surface Temperature (LST) over the study period. It is an important topic to understand urban land cover changes for better and sustainable future planning. The results are interesting and the methods used are decent. However, there are some major/minor corrections and clarifications required to be done to consider the manuscript for publication. Please see my comments below.
Broad comments
I feel that the current paper title does not reflect the main context of the paper, which is investigating the impact of urbanization on Land Surface Temperature (LST) using two methods.
The abstract could be more informative for the readers. The introductory lines are too lengthy, I am reading halfway through the abstract and I did not know anything about the main objective or the key findings of the manuscript. Include the most important/interesting quantitative results and findings.
Study area and dataset section: It should be mentioned, how many Landsat satellite imageries did you use? Furthermore, you could present a table, which shows dates of acquisition, path/row numbers, type of Landsat sensor and spatial resolution per each image.
Line 153: Section 3.2 in the Methods. I appreciate that you are providing the references for the developed and used method in your research. However, I feel that you need to slightly elaborate the methods used (3-4 lines).
Line 226: Section 3.3.2 in the Methods. You mentioned that the images were scaled down to 300 m. Do you mean 300 m of images’ spatial resolution? Why scaling the images down 10 times the resolution?
Section 4.1 in the results. I appreciate that the urban expansion and the loss of croplands in not on the focus of your paper. However, it would be interesting to mention the impact of this issue, demonstrating areas of cropland lost due to urban sprawl over the study period. Perhaps over 1 paragraph.
Table 2: What is the unit used to represent LST? It should be Kelvin but it is not mentioned within the table title.
The conclusions section is lengthy. It could be half the size and twice as effective. In addition, it should numbered as section no.5 not 4 as it stands. Some information could be moved from conclusions to the discussion section, particularly, the last paragraph regarding research limitations.
Figure 1 is confusing. It says that (b) is Colombo Metropolitan Area, while the map title says it is a map of the western province. You could include both in one unified figure to make it clearer. A suggestion here, which you could point out the location of Sri Lanka to the world or Asia.
Figure 4 is not clear enough and a bit confusing. What about using coloured lines to represent each land cover class? Moreover, (km2) on the vertical axis, the square needs to be superscripted.
Figures 7 and 8: The units representing distance and temperature on both the horizontal and the vertical axis are missing and need to be clarified.
Generally, I appreciate that the authors are not native English speakers. However, the English language level used in this manuscript is not sufficiently appropriate for publishing at this stage, and needs to undertake major English language revision.
Specific comments
Line 74: Urban land cover changes should be written as (ULCC).
Line 89: In the study area section. 699km2, square should be superscripted.
Line 130: Section 3.1 in the Methods. Urban land cover changes should be ULCC.
Line 187: Section 3.3 in the Methods. ARCGIS should be written as ArcGIS.
Line 210: Section 3.3.1 in the Methods. ARCGIS should be written as ArcGIS.
Line 230: Results and discussion section is numbered as (3). It should be numbered as (4) according to the paper’s sections sequence. Furthermore, there should be two separated sections for results and discussion according to publishing regulations in MDPI’s Remote Sensing.
Line 234: Section 4.1 in the results. 1896 should be 1988.
Line 276: Section 4.2 in the results. It should be ArcGIS not ARCGIS.
Lines 363-364: Section 4.3.2 in the results. Why write (City of Colombo city)?
Line 374: Section 4.3.2 in the results. Define (GIS). Abbreviations and terms should be defined at the first time they appear in the text.
Line 380: Section 4.3.3 in the results. Is it 1986 or 1988?
Author Response
Overall comments
This manuscript by Fonseka et al. explored the urban expansion from 1988 to 2016 in Colombo Metropolitan Area, Sri Lanka using multi-temporal Landsat images. Two methods have been used to evaluate the impact of rapid urbanization on the Land Surface Temperature (LST) over the study period. It is an important topic to understand urban land cover changes for better and sustainable future planning. The results are interesting and the methods used are decent. However, there are some major/minor corrections and clarifications required to be done to consider the manuscript for publication. Please see my comments below.
RESPONSE: Thank you very much for the generally positive comments and the useful suggestions. Below please find our item-to-item responses to the specific comments.
Especially, since our tracked changes using the “Track Changes” function in Microsoft Word were accepted by Elsevier Language Editing Service before the editing, we highlighted these changes in BLUE color in the revised manuscript.
Broad comments
I feel that the current paper title does not reflect the main context of the paper, which is investigating the impact of urbanization on Land Surface Temperature (LST) using two methods.
RESPONSE: The paper title was revised to “Urbanization and its impacts on land surface temperature in Colombo Metropolitan Area, Sri Lanka, from 1988 to 2016”.
The abstract could be more informative for the readers. The introductory lines are too lengthy, I am reading halfway through the abstract and I did not know anything about the main objective or the key findings of the manuscript. Include the most important/interesting quantitative results and findings.
RESPONSE: Thank you for the critical and useful comment and suggestion. We have revised the abstract to make it more clear, concise and shorter in the introductory parts. We also reorganized the main results and findings in the abstract by organizing them in the following four main points with quantitative results:
“The experimental results indicate that: 1) the urban land cover classification during the study period was conducted with satisfactory accuracy, with more than 80% for the overall accuracy and over 0.73 for the Kappa coefficient; 2) the Colombo Metropolitan Area exhibits a diffusion pattern of urban growth, especially along the west coastal line, from both the multi-buffer ring approach and the gravity model; 3) urban density was identified as having a positive relationship with LST through time; 4) there was a noticeable increase in the mean LST, of 5.24°C for water surfaces, 5.92°C for vegetation, 8.62°C for bare land, and 8.94°C for urban areas.”
Study area and dataset section: It should be mentioned, how many Landsat satellite imageries did you use? Furthermore, you could present a table, which shows dates of acquisition, path/row numbers, type of Landsat sensor and spatial resolution per each image.
RESPONSE: Thank you for this good suggestion. We have appended a table listing all the Landsat satellite images used in this study, together with their dates of acquisition, path/row numbers, sensors and spatial resolutions. Please kindly refer to the revised manuscript for more details.
Line 153: Section 3.2 in the Methods. I appreciate that you are providing the references for the developed and used method in your research. However, I feel that you need to slightly elaborate the methods used (3-4 lines).
RESPONSE: More descriptions about the methods were added with related references. Technical details about the two selected classification methods, support vector machine and random forest, were provided to briefly elaborate the methods. Please turn to the following paragraph in Section 3.2.
“Support Vector Machine (SVM) and Random Forest (RF), two of the most popular machine learning methods, were comparatively applied to the supervised classification of urban land cover [23-25]. First, the radial basis function (RBF) was selected as the kernel function to map the data onto a binary separable hyperplane in SVM, which was optimized with a cross-validation for the settings of two key parameters: Gramma in the RBF, and the penalty (C) for non-linear cases in the hyperplane [26-29]. In addition, since the original SVM is a binary classifier, multiple SVMs were needed to conduct the multi-class classification, with each classifier used to identify one land cover class. To perform this, the one-against-the-rest strategy was employed [28]. Second, RF is a decision tree based classifier, which consists of a set of decision trees, with each trained based on a randomly selected subset of the total training samples [29,30]. The final classification result of RF is a voting-based decision based on the classifications of all the decision trees. The successful application of RF depends on the optimal settings of two key parameters, the number of decision trees and the number of features that are randomly selected to split each node in the decision trees. According to our previous study, a greater number of decision trees will result in a better RF; however, the performance of RF will become stable with no significant improvement after a certain number of decision trees [29]. To achieve the optimal performance of RF, the number of decision trees was set to 100. For the number of features, this study followed the previous studies, to set this parameter as the root of the total features [31,32].”
Line 226: Section 3.3.2 in the Methods. You mentioned that the images were scaled down to 300 m. Do you mean 300 m of images’ spatial resolution? Why scaling the images down 10 times the resolution?
RESPONSE: We are very sorry for our mistakes in this section for the careless and incorrect presentation of our data processing and results analysis. Actually, we did not change the spatial resolution of the classification images, but only calculated the urbanization percentage using a moving window technique before applying the gravity model. We have revised this paragraph to clarify the technical details of the application of gravity model. Please kindly read the revised paragraph, as:
“The general concept of this step is expressed in Equation (5), by which the percentage of urbanization is calculated based on a binary image with a given window size. The window size was set at 9 × 9 pixels, for which we assume that a spatial scale of 270 m × 270 m is big enough to evaluate the urbanization percentage of a pixel. Then the urban density is further calculated based on the gravity model expressed in Equation (5), implemented using Matlab.”
Section 4.1 in the results. I appreciate that the urban expansion and the loss of croplands in not on the focus of your paper. However, it would be interesting to mention the impact of this issue, demonstrating areas of cropland lost due to urban sprawl over the study period. Perhaps over 1 paragraph.
RESPONSE: Thank you for the good suggestion. We have added a paragraph to discuss the impacts of the loss of vegetation, including mainly croplands and forest. Especially, according to your further comment below, we placed this paragraph in the newly added discussion section. Please refer to the revised manuscript for more details.
“The vegetation in the Colombo Metropolitan Area was mainly cropland and forest. However, due to the spectral similarity of the two vegetation types, it is difficult to accurately separate them with the 30-m resolution Landsat images. Generally, as demonstrated in Figure 4, the vegetation decreased from over 600 km2 to near 400 km2, with a decrease of over 33%. There are several negative impacts of this remarkable loss of vegetation. First, since these vegetation areas include a large amount of cropland, it is destructive of the local agriculture. Second, the ecosystem is also threatened, due to the significant decrease of both forest and the agricultural ecosystem, due to which the ecological services are weakened. Third, the conversion of this large area of vegetation to an urban area could contribute significantly to the urban heat island effects of the Colombo Metropolitan Area by increasing the LST.”
Table 2: What is the unit used to represent LST? It should be Kelvin but it is not mentioned within the table title.
RESPONSE: Sorry for our carelessness. Yes, it is Kelvin. We added the unit to the title of Table 2, which is now Table 3 in the revised manuscript.
The conclusions section is lengthy. It could be half the size and twice as effective. In addition, it should numbered as section no.5 not 4 as it stands. Some information could be moved from conclusions to the discussion section, particularly, the last paragraph regarding research limitations.
RESPONSE: Thank you very much for this critical and important suggestion. We have revised the conclusion section to be more concise and clear. The length was reduced to be around half of the original one. Especially, we added a new subsection in the newly added discussion section, named “Limitations of the research”, which includes some of the comments and discussion in the original conclusion section. Please kindly refer to the revised manuscript for more details.
Figure 1 is confusing. It says that (b) is Colombo Metropolitan Area, while the map title says it is a map of the western province. You could include both in one unified figure to make it clearer. A suggestion here, which you could point out the location of Sri Lanka to the world or Asia.
RESPONSE: Sorry for our mistakes. We have revised the map accordingly.
Figure 4 is not clear enough and a bit confusing. What about using coloured lines to represent each land cover class? Moreover, (km2) on the vertical axis, the square needs to be superscripted.
RESPONSE: Figure 4 was improved to be clearer, including the presentation of classes and the axis.
Figures 7 and 8: The units representing distance and temperature on both the horizontal and the vertical axis are missing and need to be clarified.
RESPONSE: The units were added in Figures 7 and 8.
Generally, I appreciate that the authors are not native English speakers. However, the English language level used in this manuscript is not sufficiently appropriate for publishing at this stage, and needs to undertake major English language revision.
RESPONSE: Thank you for the suggestion. We have asked professional English editing service to conduct extensive English editing over our revised manuscript. A certificate of the editing was also attached with our resubmission.
Specific comments
Line 74: Urban land cover changes should be written as (ULCC).
RESPONSE: It was corrected.
Line 89: In the study area section. 699km2, square should be superscripted.
RESPONSE: It was revised.
Line 130: Section 3.1 in the Methods. Urban land cover changes should be ULCC.
RESPONSE: The abbreviation of ULCC was double checked and revised in this section and throughout manuscript.
Line 187: Section 3.3 in the Methods. ARCGIS should be written as ArcGIS.
RESPONSE: It was revised.
Line 210: Section 3.3.1 in the Methods. ARCGIS should be written as ArcGIS.
RESPONSE: It was revised.
Line 230: Results and discussion section is numbered as (3). It should be numbered as (4) according to the paper’s sections sequence. Furthermore, there should be two separated sections for results and discussion according to publishing regulations in MDPI’s Remote Sensing.
RESPONSE: The numbering of sections was corrected throughout the manuscript. Furthermore, a new discussion section was added with two subsections, 1) Impacts of the loss of vegetation due to urbanization, and 2) Limitations of the research. We hope the structure of the revised manuscript is now improved for the better presentation and understanding of readers.
Line 234: Section 4.1 in the results. 1896 should be 1988.
RESPONSE: It was corrected.
Line 276: Section 4.2 in the results. It should be ArcGIS not ARCGIS.
RESPONSE: It was corrected.
Lines 363-364: Section 4.3.2 in the results. Why write (City of Colombo city)?
RESPONSE: It was corrected. It should be “Colombo”.
Line 374: Section 4.3.2 in the results. Define (GIS). Abbreviations and terms should be defined at the first time they appear in the text.
RESPONSE: It was revised to “Geographic Information System (GIS)”.
Line 380: Section 4.3.3 in the results. Is it 1986 or 1988?
RESPONSE: It was corrected to 1988. Similar mistakes were also corrected throughout the manuscript.
Reviewer 2 Report
The manuscript utilises existing methods to examine the impact of urban land cover change on land surface temperature in Colombo, Sri Lanka. Although authors have emphasised that they provided a comprehensive methodology, I can find that there remain huge methodological flaws in the paper. Moreover, they stated that urban land cover change usually done on temperate region but I can show hundreds of works on Dhaka, Mumbai, Calcutta, Delhi and so on. The title of the work does not reflect its content, ie. nowhere in the title you can find that they want to see the impact of ULCC on LST. Why are you saying 'the most impressive' human activities given that urban growth leads to a number of negative impacts on the environment. You are saying time-series data but Landsat data that are used in this work represents 3-4 year intervals. My specific comments are:
(1) Objectives are very unclear and does not lead to useful conclusions
(2) SVM is a kind of supervised technique but as such machine learning. I am wondering if authors can use random forest algorithm and compare with SVM outputs. This could make this work interesting
(3) A table showing Landsat data, year of acquisition etc could be useful. If you were used Landsat ETM+ after 2003 then SLC problem must be addressed, nothing is mentioned in this version
(4) Should be 'MatLab" and ArcGIS throughout the manuscript
(5) Did you use single channel method to retrieve LST, should clarify this
(6) What was the point of downscaling 30 m to 300m, no rational given. Also, line 225-229: why you had to do this since your inputs are binary image? Does normalisation was essential, I doubt
(7) Fig. 7 shows opposite trends how are you validating this?
(8) What Table 3 refers to? Nothing is clear from here. Is this correlation is based on LST and Gravity or multi-buffer ring?
(9) Readers can be greatly benefited to have a discussion section, unfortunately it is not present. I hope authors can get related articles on Dhaka, Mumbai, Delhi etc to elucidate pattern and processes of ULC in Colombo. Then they can cross check similarity and dissimilarity on the pattern and processes of ULC in South Asia. This is completely missing
Author Response
The manuscript utilizes existing methods to examine the impact of urban land cover change on land surface temperature in Colombo, Sri Lanka. Although authors have emphasised that they provided a comprehensive methodology, I can find that there remain huge methodological flaws in the paper. Moreover, they stated that urban land cover change usually done on temperate region but I can show hundreds of works on Dhaka, Mumbai, Calcutta, Delhi and so on. The title of the work does not reflect its content, ie. nowhere in the title you can find that they want to see the impact of ULCC on LST. Why are you saying 'the most impressive' human activities given that urban growth leads to a number of negative impacts on the environment. You are saying time-series data but Landsat data that are used in this work represents 3-4 year intervals. My specific comments are:
RESPONSE: Thank you very much for the critical and useful comments. We have made several major revisions including the abstract and the whole structure of the manuscript. Major revisions were applied to all literature review in the introduction section, the methods, results and discussion sections. However, since Colombo is located in a tropical region that has experienced very intensive clouds contaminations in the past three decades, it is really rather difficult to get clear sky satellite images from the free accessible Landsat database. Consequently, we were only able to obtain the relatively good satellite data with 3-4 years intervals. In this study, we tried to investigate the urbanization process of Colombo since it is becoming a significant metropolitan in Sri Lanka and its urbanization impacts are occurring to the local ecology, agriculture, environment, and the human society as well. We hope this work will: 1) extend the application time-series satellite data to the intensively rainy and cloudy tropical metropolitans, and 2) provide a scientific reference of the urbanization procedures with its impacts to the local government for their policy of urban planning and management. Please kindly refer to the following point-by-point responses for the details of our revisions.
Especially, since our tracked changes using the “Track Changes” function in Microsoft Word were accepted by Elsevier Language Editing Service before the editing, we highlighted these changes in BLUE color in the revised manuscript.
(1) Objectives are very unclear and does not lead to useful conclusions
RESPONSE: Thank you for the critical comment. We rewrote the abstract to better outline the objectives and summarize the conclusions of the study. The revised abstract is:
“Urbanization has become one of the most important human activities to modify the Earth’s land surfaces, and its impacts on tropical and subtropical cities (e.g. in South/Southeast Asia) are not fully understood. One of these cities, Colombo, the capital of Sri Lanka, has been urbanized for about 2,000 years, due to its strategic position on the East-West sea trade routes. This study aims to investigate the characteristics of urban expansion and its impacts on land surface temperature in Colombo from 1988 to 2016, using a time-series of Landsat images. Urban land cover changes (ULCC) were derived from the time-series satellite images with the assistance of machine learning methods. Urban density was selected as a measure of urbanization, derived from both the multi-buffer ring method and a gravity model, which were comparatively adopted to evaluate the impacts of ULCC on the changes of land surface temperature (LST) over the study period. The experimental results indicate that: 1) the urban land cover classification during the study period was conducted with satisfactory accuracy, with more than 80% for the overall accuracy and over 0.73 for the Kappa coefficient; 2) the Colombo Metropolitan Area exhibits a diffusion pattern of urban growth, especially along the west coastal line, from both the multi-buffer ring approach and the gravity model; 3) urban density was identified as having a positive relationship with LST through time; 4) there was a noticeable increase in the mean LST, of 5.24°C for water surfaces, 5.92°C for vegetation, 8.62°C for bare land, and 8.94°C for urban areas. The results provide a scientific reference for policy makers and urban planners working towards a healthy and sustainable Colombo Metropolitan Area.”
(2) SVM is a kind of supervised technique but as such machine learning. I am wondering if authors can use random forest algorithm and compare with SVM outputs. This could make this work interesting
RESPONSE: Thank you very much for the useful suggestion. We have added random forest algorithm to compare with the SVM outputs. The urban land cover classification results in Colombo from 1988 to 2016 are shown in Figure 3 using SVM and Figure 4 using RF. As reported in our previous study [38], the performance of classification using SVM and RF is comparable, and SVM often slightly outperforms RF, while RF often has better efficiency since it is based on decision trees, the classification in this study also matched our previous study. From Figures 3 and 4, the classification results were mostly consistent; while RF appears a little bit over estimation of urban areas in the year 2013 and 2016. However, RF used much less time compared with SVM during our experiments. For instance, it took more than ten minutes to classify one image for SVM, but it took less than one minute to classify one image for RF. Nevertheless, considering that this study focused more on the accuracy than on the computational time, we adopted the results from SVM for our further analysis.
(3) A table showing Landsat data, year of acquisition etc could be useful. If you were used Landsat ETM+ after 2003 then SLC problem must be addressed, nothing is mentioned in this version
RESPONSE: We have appended a table listing all the Landsat satellite images used in this study, together with their dates of acquisition, path/row numbers, sensors and spatial resolutions. We did not use the Landsat ETM+ after 2003 due to the SLC problem. Please kindly refer to the revised manuscript for more details.
(4) Should be 'MatLab" and ArcGIS throughout the manuscript
RESPONSE: The two terms were revised throughout the manuscript.
(5) Did you use single channel method to retrieve LST, should clarify this
RESPONSE: We are sorry for not presenting this information clearly. We have added related information in the revised manuscript. Actually, not only single channel was used to retrieve LST. According to the detailed methods of LST estimation described in Section 3.2, the calculation of LST needs the estimation of at-sensor brightness temperature (Kelvin) and the emissivity. The at-sensor brightness temperature can be calculated from the thermal infrared channel, while the emissivity should be derived from visible and near infrared channels. In this study, we adopted the most frequently used method to estimate the emissivity based on the normalized difference vegetation index (NDVI), which needs the red and near infrared channels.
(6) What was the point of downscaling 30 m to 300m, no rational given. Also, line 225-229: why you had to do this since your inputs are binary image? Does normalisation was essential, I doubt
RESPONSE: We are very sorry for our mistakes in this section for the careless and incorrect presentation of our data processing and results analysis. Actually, we did not change the spatial resolution of the classification images, but only calculated the urbanization percentage using a moving window technique before applying the gravity model. We have revised this paragraph to clarify the technical details of the application of gravity model. Please kindly read the revised paragraph, as:
“The general concept of this step is expressed in Equation (5), by which the percentage of urbanization is calculated based on a binary image with a given window size. The window size was set at 9 × 9 pixels, for which we assume that a spatial scale of 270 m × 270 m is big enough to evaluate the urbanization percentage of a pixel. Then the urban density is further calculated based on the gravity model expressed in Equation (5), implemented using Matlab. The urban density results were normalized into a scale from 0 to 1, after which a correlation analysis was conducted with the LST values. Normalization was conducted mainly for a better visualization of the urban density in Section 4.3.2. Since it is a linear normalization, it does not change the pattern of the urban density.”
(7) Fig. 7 shows opposite trends how are you validating this?
RESPONSE: As shown in the Figure 7, both urban density and mean temperature had decreasing trend from the city centre although the slope of decreasing is not in equal rates. It was clear that in both diagrams, it had fluctuations in some buffer areas which can be attributed to the trend of the urban expansion of Colombo that is not radial. The reason why the changing patterns between the urban density and LST are not exactly the same is that urban density does not influence the LST directly, but through the changes of land surface emissivity, which in turn changes the LST in a non-linear way, as shown in Equation (3). However, the exact relationship between urban density and LST is difficult to validate through in-situ measurement, because the emissivity can be influenced by a number of factors such as the composition of all land cover classes within every pixel. Therefore, a qualitative analysis of the relationship between urban density and LST is demonstrated in Figure 7 and Figure 8. The correlation is also employed to quantify this complicated relationship in Table 3.
(8) What Table 3 refers to? Nothing is clear from here. Is this correlation is based on LST and Gravity or multi-buffer ring?
RESPONSE: We are sorry for our misleading and unclear presentation of Table 3 (Table 4 in the revised manuscript). Yes, it should be the correlation coefficients between average urban densities and mean LST derived from multi-buffer ring and gravity model. We have corrected the title and added related explanations in the text.
(9) Readers can be greatly benefited to have a discussion section, unfortunately it is not present. I hope authors can get related articles on Dhaka, Mumbai, Delhi etc to elucidate pattern and processes of ULC in Colombo. Then they can cross check similarity and dissimilarity on the pattern and processes of ULC in South Asia. This is completely missing.
RESPONSE: Thank you for this critical and useful suggestion. In the revised manuscript, we added a new discussion section. Several issues were discussed in this section: “Urban growth is common in many South Asia countries, while their growing process and patterns are different due to different urban planning and administration policies as well as their different urban landscape. Many South Asia cities were studied using satellite images and dramatic urbanization was often reported in these studies [9-11,13,53]. The urbanization amplitude of most major cities, such as Mumbai, Delhi, Hanoi and Dhaka, was reported to be similar to that in Colombo of this study according to their urban land cover changes [7,9,10]. The relationship between LST and urban land use or land cover types were investigated with similar correlation [9,12,13,54]. However, the usage of satellite images was different in different studies. Studies using satellite images at two or three different dates were the most common [7-10]. All available clear-sky images within several years were conducted in Bangkok and Mumbai [11,12]. Since South Asia is located in a tropical and subtropical region, clouds contamination is a common problem for the application of optical remote sensing for urban observations, which is also indicated in most existing studies. Therefore, multi-source satellite images, especially the incorporation of microwave radar data, were recommended for future studies in South Asia.”
Round 2
Reviewer 1 Report
I would like to thank the authors for revising the manuscript according to the reviewers' comments. Generally, this revised version of the manuscript has been significantly improved over the last iteration of the manuscript.
P.S: I think as a matter of courtesy, it would be best if you could include thanks to the reviewers for their constructive comments and suggestions in the acknowledgments section.
Reviewer 2 Report
Even though the revised version has been improved, I still feel that this manuscript does not offer any novelty. What they have presented here is already known, except application to Colombo. I hoped that they could elevate the revision compare to original submission but it did not. This manuscript should go to a regional journal
