Using Dual Spatial Clustering Models for Urban Fringe Areas Extraction Based on Night-time Light Data: Comparison of NPP/VIIRS, Luojia 1-01, and NASA’s Black Marble

Round 1

Reviewer 1 Report

The manuscript presents a comprehensive comparison of various algorithms in the field of delineating urban fringe areas. Overall, this paper is well-organised and focused on its theme, and the methodology appears promising. Nonetheless, some revisions are necessary prior to the formal acceptance of this paper.

1. Overall, the manuscript could benefit from more detailed description of the algorithms being compared. While the authors mention the comparative algorithms, it would be beneficial to provide a succinct overview of the key principles, advantages, and limitations for each.

2. Line 103: Please include the definition of the phrase 'in the presence of heterogeneity and noise.'

3. Figure 1: It is recommended to replace the DEM with land use classification map; 'Jiangye' should be corrected to 'Jianye.'

4. Figure 7: Some illustration of this figure is required in Section 4.

5. Furthermore, the inconsistent use of specialised jargon, such as “MK-means” or “MK-Means,” and “NPP/VIIRS” or “NPP,” should be addressed throughout the manuscript.

No

Author Response

We really appreciate your valuable comments and suggestions. In this revision, we have made the following revisions to address your concerns (the revised text in red font):

Reviewer #1：

The manuscript presents a comprehensive comparison of various algorithms in the field of delineating urban fringe areas. Overall, this paper is well-organised and focused on its theme, and the methodology appears promising. Nonetheless, some revisions are necessary prior to the formal acceptance of this paper.

Response:

We really appreciate this valuable comment and are convinced that our paper can be improved according to the specific suggestions.

1．Overall, the manuscript could benefit from more detailed description of the algorithms being compared. While the authors mention the comparative algorithms, it would be beneficial to provide a succinct overview of the key principles, advantages, and limitations for each.

Response:

Thanks for your comments. Following your suggestion, we have provided a succinct overview of the key principles, advantages, and limitations for each algorithm in the submitted version of the manuscript.

• The key principle of MK-Means can be found in Section 3.2.1. The advantage and limitation for this algorithm was presented in Section 1. (See Lines 91-98 in the submitted version of the manuscript)

“Although the MK-Means algorithm is an extension of the K-means algorithm that incorporates at-tribute metrics to broaden the focus on the spatial object's attribute distance, aiming to consider both the heterogeneity of spatial positions and the similarity of attributes, the analysis results continue to be challenged by issues such as varying density, arbitrary shapes, and spatial noise problems, which cannot be overlooked.”

• The key principle of DBSC can be found in Section 3.2.2. The advantage and limitation for this algorithm was presented in Section 1. (See Lines 100-105 in the submitted version of the manuscript)

“The DBSC algorithm has proven to be efficient and applicable in identifying clusters with irregular shapes and varying densities. It has been successfully used for urban element identification and urban spatial structure analysis [35-36]. However, DBSC algorithm results heavily depend on the global shared non-spatial attribute threshold, failing to capture the inhomogeneous structure of the dataset.”

• The key principle of DSC can be found in Section 3.2.3. The advantage and limitation for this algorithm was presented in Section 1. (See Lines 105-113 in the submitted version of the manuscript)

“DSC can handle both spatial proximity and attribute similarity in the presence of hetero-geneity (i.e., the difference of observations in attribute distribution is homogeneous within each cluster but inhomogeneous between clusters) and noise (i.e., attribute values of spa-tial objects are significantly different from those of other objects in its spatial neighborhood). Nonetheless, the DSC algorithm may result in over-segment clusters when dealing with mixed inhomogeneous and homogeneous dataset structures.”

2. Line 103: Please include the definition of the phrase 'in the presence of heterogeneity and noise.'

Thanks for your comments. Following your suggestion, we have added the definition of the phrase 'in the presence of heterogeneity and noise.' (See Lines 106-108 in the submitted version of the manuscript)

“DSC can handle both spatial proximity and attribute similarity in the presence of hetero-geneity (i.e., the difference of observations in attribute distribution is homogeneous within each cluster but inhomogeneous between clusters) and noise (i.e., attribute values of spa-tial objects are significantly different from those of other objects in its spatial neighbor-hood).”

3. Figure 1: It is recommended to replace the DEM with land use classification map; 'Jiangye' should be corrected to 'Jianye.'

Thanks for your comments. Following your suggestion, we have revised Figure 1 in the submitted version of the manuscript.

4. Figure 7: Some illustration of this figure is required in Section 4.

Thanks for your comments. Following your suggestion, we have added illustration of Figure 7 in the submitted version of the manuscript. (See Lines 436-438 in the submitted version of the manuscript)

“As a result, 329, 339, and 445 mutation points were identified in NPP/VIIRS, Luojia 1-01, and NASA’s Black Marble data, respectively, with many intersection points (Figure 7).”

5. Furthermore, the inconsistent use of specialised jargon, such as “MK-means” or “MK-Means,” and “NPP/VIIRS” or “NPP,” should be addressed throughout the manuscript.

Thanks for your comments. Following your suggestion, we have revised this grammatical mistake in the submitted version of the manuscript.

Finally, we are grateful for your comments and suggestions. The comments are all valuable and have helped us to revise and improve our manuscript, as well as provided important guidance for our research. We would be glad to respond to any further questions and comments.

Thank you again.

Author Response File: Author Response.pdf

Reviewer 2 Report

In the Study Area and Data section, the paper states that all the data were reprojected and resampled to 500m x 500m. What method is used for the resampling? Presumably nearest neighbor for NPP/VIIRS and Black Marble, but what about for Luojia? And were the quality flags for the Black Marble data used at all?

Figure 2 displays the resampled data. What are the units of measurement? I know for a fact that the Black Marble data is in nW·cm-2 ·sr-1 so how has the data been scaled from 0-255? Similar question for NPP/VIIRS and Luojia. Why wasn’t a common scaling used?

Line 114 states that the NLI index was used, but this term is not defined until line 162. And that definition is puzzling:  “NLI is quantified by obtaining the digital number of pixels from the images.” What does that even mean, and again, how and why does it differ for each data source?

The paragraph from lines 159-172 needs work – what is this new grid cell? The data was already resampled to 500m x 500m.

Line 499 needs editing.

What is meant by "Black Marble data with medium and high spatial resolution" line 577? All of the VNP46 products are at 15 arcsec resolution, which is approximately 500m at the equator.

Author Response

Response to the reviewers

We really appreciate your valuable comments and suggestions. In this revision, we have made the following revisions to address your concerns (the revised text in red font):

Reviewer #2：

1. In the Study Area and Data section, the paper states that all the data were reprojected and resampled to 500m x 500m. What method is used for the resampling? Presumably nearest neighbor for NPP/VIIRS and Black Marble, but what about for Luojia? And were the quality flags for the Black Marble data used at all?

Response:

Thanks for your valuable suggestions. Following your suggestion, we have made the following revisions to address your concerns:

• For consistent analysis and comparison across various data sources, NPP-VIIRS and NASA's Black Marble data were resampled to a spatial resolution of 500 meters × 500 meters (cell size) using the nearest neighbor method within the ArcGIS “Resample” tool, while Luojia 1-01 data were resampled to the same resolution using the aggregating rule within the ArcGIS “Aggregate” tool. (See Lines 163-168 in the submitted version of the manuscript)
• The Black Marble data preprocessing in our study has been carried out according to Zheng’s work (https://github.com/qmzheng09work/NTL‐VIIRS‐BlackMarbleProduct), the data layer covers all the quality flags, including 0(high quality),1(high quality),2(low quality),255(missing data). (See Lines 161-162 in the submitted version of the manuscript)

2. Figure 2 displays the resampled data. What are the units of measurement? I know for a fact that the Black Marble data is in nW·cm-2 ·sr-1 so how has the data been scaled from 0-255? Similar question for NPP/VIIRS and Luojia. Why wasn’t a common scaling used?

Response:

Thanks for your valuable suggestions.

Firstly, we would like to express our sincere apologies to you for any confusion that may have arisen from our manuscript. We recognize that a type mistake occurred in our work. Actually, the unit of all resampled data in Figure 2 is nW·cm-2 ·sr-1, not “DN”. To reduce the effect of light saturation, a radiometric correction for Luojia1-01 NTL has been implemented using the formula  . We then converted the unit of Luojia 1-01 radiance to , which matches the unit of VIIRS datasets. (See Lines 147-149 in the submitted version of the manuscript)

Secondly, as previous mentioned, the Black Marble data preprocessing has been carried out according to Zheng’s work (https://github.com/qmzheng09work/NTL‐VIIRS‐BlackMarbleProduct). In this data preprocessing, the actual radiance values of pixels used in Figure 2(c) are derived by multiplying the original data by 0.1, rather than being scaled to a range of 0-255. Simultaneously, we also identified very few missing values in the downloaded data. However, we did not address these missing values because they had no impact on the assessment of our clustering results and urban fringe extraction outcomes.

In summary, there was an error in our labeling, and the correct units for all resampled data in Figure 2 are nW·cm-2 ·sr-1. (See Figure 2 in the submitted version of the manuscript)

3. Line 114 states that the NLI index was used, but this term is not defined until line 162. And that definition is puzzling: “NLI is quantified by obtaining the digital number of pixels from the images.” What does that even mean, and again, how and why does it differ for each data source?

Response:

Thanks for your valuable suggestions. The NLI index represents the nighttime light intensity (See Line 74 in the submitted version of the manuscript). Indeed, the NLI index in this study corresponds to the radiance value of each pixel derived from the NTL images. We have revised the definition of NLI index in the submitted version of the manuscript. (See Line 174-175 in the submitted version of the manuscript)

4. The paragraph from lines 159-172 needs work – what is this new grid cell? The data was already resampled to 500m x 500m.

Response:

Thanks for your valuable suggestions. In our work, we converted the nighttime light pixels to grid square cells and assigned the radiance value of each pixel to each grid cell. Indeed, this is essentially a preprocessing step. The primary purpose is to facilitate the subsequent process of detecting mutation points using SCWT. In a previous study [31], a city center was used as the starting point and an interval of one degree was selected to obtain 360 sampling lines. Furthermore, the center of the outermost grid of the entire study area was set as the end point. This method ensured that each grid cell was calculated. The spatial urbanization sequence signal of each sample line was then constructed by recording the urbanization indices (NLI index in this study) of the grids intersecting the sample line from the city center to the periphery. Next, the mutation points of each signal (i.e., maximum and minimum values) were identified using the SCWT method with the optimal scale and mapped to determine the spatial mutation points of urbanization in study area. In the submitted version of the manuscript, we have revised this paragraph (See Line 181-184 in the submitted version of the manuscript) and included additional details pertaining to the relationship between grid cells and mutation points using SCWT (See Line 201-211 in the submitted version of the manuscript).

5. Line 499 needs editing.

Response:

Thanks for your valuable suggestions. we have revised this sentence in the submitted version of the manuscript. (See Line 515-516 in the submitted version of the manuscript)

6. What is meant by "Black Marble data with medium and high spatial resolution" line 577? All of the VNP46 products are at 15 arcsec resolution, which is approximately 500m at the equator.

Response:

Thanks for your valuable suggestions. We have revised this sentence into “NASA's Black Marble data can provide more detailed information about intracity NTL variations, which was previously less achievable with VIIRS data [10]” in the submitted version of the manuscript. (See Line 559 in the submitted version of the manuscript)

Finally, we are grateful for your comments and suggestions. The comments are all valuable and have helped us to revise and improve our manuscript, as well as provided important guidance for our research. We would be glad to respond to any further questions and comments.

Thank you again.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper entitled “Using Dual Spatial Clustering models for Urban fringe Areas Extraction Based on Nighttime Light Data: Comparison of NPP/VIIRS, Luojia 1-01 and NASA's Black Marble” presents an interesting research which fits to the scope of the journal. In my opinion the submitted version of the manuscript is suitable for publication.

The paper addresses an important research topic. Comparison of three methods (MK-means, DBSC, and DSC) for nightlight data analysis presents possible outcomes to obtain in urban fringe analysis. Findings might be useful in land use planning field. Obtained conclusions are consistent with the evidence and arguments presented and do they address the main research question. References used in this research are appropriate. I do not have any suggestions to improve the paper. In my opinion it can be published as it is presented in the current version.

Author Response

Reviewer #3：

The paper entitled “Using Dual Spatial Clustering models for Urban fringe Areas Extraction Based on Nighttime Light Data: Comparison of NPP/VIIRS, Luojia 1-01 and NASA's Black Marble” presents an interesting research which fits to the scope of the journal. In my opinion the submitted version of the manuscript is suitable for publication.

The paper addresses an important research topic. Comparison of three methods (MK-means, DBSC, and DSC) for nightlight data analysis presents possible outcomes to obtain in urban fringe analysis. Findings might be useful in land use planning field. Obtained conclusions are consistent with the evidence and arguments presented and do they address the main research question. References used in this research are appropriate. I do not have any suggestions to improve the paper. In my opinion it can be published as it is presented in the current version.

Response:

We sincerely appreciate your recognition of our work.

Author Response File: Author Response.pdf