Using Nighttime Light Data and POI Big Data to Detect the Urban Centers of Hangzhou
Round 1
Reviewer 1 Report
Dear Authors,
The paper on “Using Nighttime Light Data and POI Big Data to Detect the Urban Centers of Hangzhou” addresses a very relevant and challenging problem, the definition of centres in rapidly growing Chinese cities. The definition of what the centre of city is, is not simple in China. The proposed methodology combines night time light with POI data to define centres and compares the results with population concentration and the centres defined in local planning documents. The study is innovative in terms of showing a data driven approach to help the definition of centres in particular in the context polycentric urban structures. However, the paper would require major improvements:
- First of all, the concept of what is a centre is not defined. Therefore the analysis is measuring something what is not clear. For example, in your results – your centres are large - common definition of centre is a point or much smaller areas?
- It is also unclear how do you define population (used for assessing the accuracy). This seems simple: but are you talking about the nighttime population (this is normally provided by census population data) or daytime population (or both). For example, you mention (line 412) that the “the airport has a dense and active population” – in a common understanding (census) an airport is not having a population – of course, it has a lot of “daytime population”. Therefore, I am not certain whether your “accuracy assessment” – comparing your results with population concentration makes a lot of sense. Centres are normally places of high concentration and diversity of urban functions and not residential places. If your population statistics would cover the day-time population this would be different (but I doubt that the local authorities have day time population counts).
- The abstract misses the problem addressed - I know from own experiences that the definition of centres in Chinese cities is a complex problem - this should be mentioned in the abstract. Please add this as the connection between the two sentences that are presently not linked in line 15.
- Line 38/39: is this true that you will have always one main centre – there might be urban agglomerations with several main centres?
- Section 51-59: references are missing!
- Line 67: ISS images provide even more detail about the urban structure, e.g., https://tandfonline.com/doi/full/10.1080/22797254.2019.1617642
- Line 68-70: the sentence is not clear – what do you mean by “reliability of district shapes” – how can be a shape reliable?
- Line 81: “Most urban studies performed by domestic and foreign scholars”- please remove “domestic and foreign” -> why would this be important?
- Line 106: please add the growth rate.
- Line 120: it is not clear whether you are using one image (May 2018) or several images?
- Figure 2: caption -> please convert the area to ha.
- Line 137etc: the entire section 2.2.3 needs a revision. First, it is not a good practice to start a section with bullets. Second, I cannot find the “Taile map downloader” -> add at least the website. Third, what do you mean by Level 17? Forth, the nature of your population data is not clear – one point per person? Such data are normally aggregated at administrative units or census tracts!
- Line 159-160: This sentence is not clear – please revise!
- Line 160: what is a “Google remote sensing map”? Do you mean Google Earth image?
- Line 176: do you have values below zero – why? And in general, for the entire section – there is literature about the correction of NTL images.
- Section 187-196: need a revision: The sequence is odd – you state that you use multi-resolution segmentation and later you list several other segmentation algorithms. I would also shorten the section. Please also explain better why to use segments – e.g. refer to https://www.sciencedirect.com/science/article/pii/S0924271613002220
- Section 3.2.1. and 3.2.3. should be combined as they are both on the scale factor. Furthermore, why did you not use ESP – this is the standard method to define the scale (e.g., https://www.tandfonline.com/doi/full/10.1080/13658810903174803).
- Line 232: it seems you used also population data to define centres and later you assessment the accuracy with the same data – this is violating the independence of validation from training or building the model.
- Line 236-7: references are missing!
- Line 239: this is within a defined neighborhood!
- Equation 7: this is the z-score!
- Equation 9: please check – to me it seems wrong!
- Line 250 etc: be more specific - what are typical critical values! The final section of this paragraph is not clear at all.
- Line 292 – why was 1.96 used?
- Figure 11: what is LMI? I could not find this in your text!
- Figure 13: what are both?
- Table 1: These results do not make sense! Most of the area is not a centre thus the OA is very high (and the kappa very low) - but what does this tell you?
- Line 445: How can a report read a plan?
- Line 455: Why the “approximate locations of the centers”? And not the location…?
- Figure 14: Terminology is not clear – what is the different between subcentre and main subcentre?
- Line 492: Why should a centre have clear boundaries” - the concept of a centre is not crisp!
- The text would require a language review and please correct the missing spaces between words (e.g., line 53, 161).
-
-
Author Response
Response to Reviewer 1 Comments
Point 1: First of all, the concept of what is a centre is not defined. Therefore the analysis is measuring something what is not clear. For example, in your results – your centres are large - common definition of centre is a point or much smaller areas?
Response 1: Thank you for your comments. We defined the main center in 3.3.1 and subcenter in 3.3.2. The study is aimed to measure a region of city centers not points. Based on the field investigation, Hangzhou is a city expanding and developing on the basis of the old city. At present, the old city still has a high population density and numerous urban facilities. Besides, the extent of the old city overlaps roughly with one of the main centers. The other main center covers the new town, which is the focus of the master plan. So the range of main center detected in this paper is understandable.
Point 2: It is also unclear how do you define population (used for assessing the accuracy). This seems simple: but are you talking about the nighttime population (this is normally provided by census population data) or daytime population (or both). For example, you mention (line 412) that the “the airport has a dense and active population” – in a common understanding (census) an airport is not having a population – of course, it has a lot of “daytime population”. Therefore, I am not certain whether your “accuracy assessment” – comparing your results with population concentration makes a lot of sense. Centres are normally places of high concentration and diversity of urban functions and not residential places. If your population statistics would cover the day-time population this would be different (but I doubt that the local authorities have day time population counts).
Response 2: Thank you for your comments. In Hangzhou, human activities at night are similar with those in the daytime. This is because the residence zones are not separated. So we did not make a distinction between the nighttime population and the daytime population.
Point 3: The abstract misses the problem addressed - I know from own experiences that the definition of centres in Chinese cities is a complex problem - this should be mentioned in the abstract. Please add this as the connection between the two sentences that are presently not linked in line 15.
Response 3: Thank you for your comments. We revised as follow: In recent years, China's large cities have put forward a multi-center urban space structure. However, the definition and identification city centers is complex.
Point 4: Line 38/39: is this true that you will have always one main centre – there might be urban agglomerations with several main centres?
Response 4: Thank you for your comments. We revised as follow: For the latter case, a city contains multiple urban centers, including one or more main urban center(s) and several sub-urban centers.
Point 5: Section 51-59: references are missing!
Response 5: Thank you for your comments. We revised it.
Point 6: Line 67: ISS images provide even more detail about the urban structure, e.g., https://tandfonline.com/doi/full/10.1080/22797254.2019.1617642
Response 6: Thank you for your comments. Photographs taken at night from the International Space Station (ISS) have finer spatial and spectral resolution than existing nocturnal observing satellites. However, it might not stand for a better result for our experiment. For example, Sharolyn et al. (2010) suggested that ‘exclusive use of nighttime images with finer spatial and spectral resolution will not necessarily improve our ability to use nighttime imagery for modelling traditional representations of population. However, analysis of the spatial patterns of error indicates that finer resolution imagery may be a good proxy of conceptualizations of population density that account for human spatial behaviour.’ So we think future research may study on the imagery such as the ISS photographs using the proposed method.
Reference: Anderson, S.J.; Tuttle, B.T.; Powell, R.L.; Sutton, P.C. Characterizing relationships between population density and nighttime imagery for Denver, Colorado: issues of scale and representation. Int. J. Remote Sens. 2010, 31, 5733-46. [doi10.1080/01431161.2010.496798]
Point 7: Line 68-70: the sentence is not clear – what do you mean by “reliability of district shapes” – how can be a shape reliable?
Response 7: Thank you for your comments. We would like to express the objectivity of the nighttime light imagery and mean it can accurately reflect the shape of physical facilities in the area. We revised as : Although nighttime light imagery has relatively high spatial stability and can guarantee the reliability of district shapesobjectivity,…
Point 8: Line 81: “Most urban studies performed by domestic and foreign scholars”- please remove “domestic and foreign” -> why would this be important?
Response 8: Thank you for your comments. We revised as follow: Most urban studies performed by scholars
Point 9: Line 106: please add the growth rate.
Response 9: Thank you for your comments. We revised as follow: In recent years, the urban land use of Hangzhou has grown rapidly, with 20% avrage growth rate from 2000 to 2016,…
Point 10: Line 120: it is not clear whether you are using one image (May 2018) or several images?
Response 10: Thank you for your comments and we are sorry for confused you. The VIIRS data we used is monthly average data in May 2018. It is one composite image that made from several daily images in a month.
We described this in 2.2.1 as:The nighttime light (NTL) data used in our study were the NPP-VIIRS monthly average radiance composite images from the VIIRS Day/Night Band (DNB), acquired in May 2018.
And we added more details in 2.2.1 as: It was one composite image that made from several daily images in a month.
The relative expression in 3.1.1 changed as: the monthly average composite nighttime lighting image.
Point 11: Figure 2: caption -> please convert the area to ha.
Response 11: Thank you for your comments. We revised it.
Point 12: Line 137etc: the entire section 2.2.3 needs a revision. First, it is not a good practice to start a section with bullets. Second, I cannot find the “Taile map downloader” -> add at least the website. Third, what do you mean by Level 17? Forth, the nature of your population data is not clear – one point per person? Such data are normally aggregated at administrative units or census tracts!
Response 12: Thank you for your comments. We revised the organization in this part as follow:
Google satellite image and government statistic population data were also applied in this study. A Google satellite image2 was used as auxiliary data and was acquired by the ArcTiler (http://www.arctiler.com/p_downloader.html), with a spatial resolution of 1.19 m. ArcTiler provided several types of remote sensing data and products. Each type of data had different resolution levels from which users could choose. We input the boundary shape file and chose the WGS84 projection. The downloaded data need no further processing.
The population data in point type were collected by the local government in 2017, each record of it represents one person with his or her location information in the research area. They were plotted by the coordinates onto a gridded map with the WGS84 projection and 500 m resolution, for consistency with the NTL data.
and here are some explanations:
- Taile map downloader is called ArcTiler (http://www.arctiler.com/p_downloader.html). It is a map download application. We corrected the expression of its name in the paper.
- ArcTiler provides different resolution levels of map for users to choose (the following figure). We considered the file size and the efficiency of image loading and then chose the Level 17 image with the resolution of 1.19m.
We deleted the Level 17 in the manuscript.
- The population data in point type were collected by the local government in 2017, each record of it represents one person with his or her location information in the research area.
Point 13: Line 159-160: This sentence is not clear – please revise!
Response 13: Thank you for your comments. We revised as follow: The nighttime light image of the study area was extracted from a global image using the administrative boundary vector.
Point 14: Line 160: what is a “Google remote sensing map”? Do you mean Google Earth image?
Response 14: Thank you for your comments and we are sorry for confused you. One of the auxiliary data we used is Google satellite image. We revised the expression. It is the satellite image in Google Map. The satellite images provided by Google Map are basically the same as those provided by Google Earth.
Point 15: Line 176: do you have values below zero – why? And in general, for the entire section – there is literature about the correction of NTL images.
Response 15: Thank you for your comments. We used the water pixels value (0) to filter the pixels with values under zero but the VIIRS image of the study area did not have values below zero.
Point 16: Section 187-196: need a revision: The sequence is odd – you state that you use multi-resolution segmentation and later you list several other segmentation algorithms. I would also shorten the section. Please also explain better why to use segments – e.g. refer to https://www.sciencedirect.com/science/article/pii/S0924271613002220
Response 16: Thank you for your comments. We revised this part as follow:
To achieve multi-source data fusion and clustering feature analysis, it was necessary to establish a unified spatial unit. Although there were some administrative divisions, human activities are not limited by the administrative boundaries. Therefore, the new statistical units were needed. We used object-based segmentation, by which homogeneous areas with the same spectral or texture features are divided into the same "object". Nighttime light images were collected by sensors from the real light on the earth’s surface, and the entity objects had different sizes, so the segmentation based on a certain scale could not make good use of its textural features.
The reason to build segements is the statistical units were needed but if we used administrative divisions or road net data, it might not reflect human activity aggregation as better as the nighttime light imagery.
Point 17: Section 3.2.1. and 3.2.3. should be combined as they are both on the scale factor. Furthermore, why did you not use ESP – this is the standard method to define the scale (e.g., https://www.tandfonline.com/doi/full/10.1080/13658810903174803).
Response 17: Thank you for your comments. Section 3.2.3 was based on the shape and compactness factors determined in 3.2.2.
Point 18: Line 232: it seems you used also population data to define centres and later you assessment the accuracy with the same data – this is violating the independence of validation from training or building the model.
Response 18: Thank you for your comments. We use POI data and NTL data to define centers and they just reflect human activity to some extent. So we might not use population data to define centers.
Point 19: Line 236-7: references are missing!
Response 19: Thank you for your comments. We added two references here:
Moran’s I is a common indicator reflecting spatial autocorrelation [34]. Generally, it can be divided into the global Moran's I proposed in 1948 and the Anselin Local Moran's I proposed by professor Luc Anselin from Arizona State University [35].
34. Deng, M.; Liu, Q. Spatial analysis, 2015 ed.; Surveying and mapping press: Beijing, China, 2015.
35. Su, L.Y.; Yang, X.H.; Bai, G.Y.; Li, F.L. Spatial inequality and regional difference of population birth rate in China. J. Chongqing Univ. Technol 2018, 32, 249-258.
Point 20: Line 239: this is within a defined neighborhood!
Response 20: Thank you for your comments. We revised this part as follow: The former reflects the spatial autocorrelation characteristics of all spatial units, while the latter reflects the spatial autocorrelation strength between a single spatial unit and other spatial units within a defined neighborhood.
Point 21: Equation 7: this is the z-score!
Response 21: Thank you for your comments and we are sorry for confused you. We just mean the Z score.
Point 22: Equation 9: please check – to me it seems wrong!
Response 22: Thank you for your suggestion. We corrected it to a clear form as:
.
Point 23: Line 250 etc: be more specific - what are typical critical values! The final section of this paragraph is not clear at all.
Response 23: Thank you for your comments and we are sorry for confused you. We revised this part as follow: A high positive value for Ii or Z(Ii ) (larger than 1.96) indicates that i is a statistically significant (0.05 level) spatial outlier [3]. According to the significance level of the local Moran’s I, the degree of spatial difference between each region and surrounding areas can be divided into two categories by combining the local Moran’s I with a Moran scatter diagram. The first type…
Point 24: Line 292 – why was 1.96 used?
Response 24: Thank you for your comments. We used the standard residual. The sub-center candidates are those sites with standard residuals > 1.96, implying that their POI density values are significantly higher than average at the local scale. We added this explanation there.
Point 25: Figure 11: what is LMI? I could not find this in your text!
Response 25: Thank you for your comments. LMI referes to Anselin Local Moran's I. We added it in 3.3.1.
Point 26: Figure 13: what are both?
Response 26: Thank you for your comments. We corrected as follow: The overlapped center areas were shown in green and the different center areas were shown in gray
Point 27: Table 1: These results do not make sense! Most of the area is not a centre thus the OA is very high (and the kappa very low) - but what does this tell you?
Response 27: Thank you for your comments. Table 1 is used to show the results of the combination of nighttime light data and POI data under LMI and GWR (our method) were better than other comparion experiments’ results because of the highest OA and kappa.
Point 28: Line 445: How can a report read a plan?
Response 28: Thank you for your comments and we are sorry for confused you. What we were trying to say was the Report interpreted (or read, unscrambled) the plan. We corrected this sentence as follow: The Report evaluates the implementation of various aspects of the plan, such as the transportation system, industry, and ecological environment.
Point 29: Line 455: Why the “approximate locations of the centers”? And not the location…?
Response 29: Thank you for your comments. We revised as follow: The locations of the centers are shown in Figure 14.
Point 30: Figure 14: Terminology is not clear – what is the different between subcentre and main subcentre?
Response 30: Thank you for your comments. We corrected them as the third grade center, the fourth grade center respectively.
Point 31: Line 492: Why should a centre have clear boundaries” - the concept of a centre is not crisp!
Response 31: Thank you for your comments. The centers in our results might not be entirely true or consistent with other studies, but the clear boundaries are not incompatible.
Point 32: The text would require a language review and please correct the missing spaces between words (e.g., line 53, 161).
Response 32: Thank you for your comments and we are sorry for confused you. We conducted language editing again and corrected the missing spaces.
Reviewer 2 Report
2.2.2 POI DATA
Baidu map? maybe is https://en.wikipedia.org/wiki/Baidu_Maps ?
It is better to indicate what it is.
Can the POI information be downloaded? What process have you adopted?
Also how many POIs are in May 2018 on Baidu?
And what is Taile map downloader?
This part must be well explained to understand the data source, the completeness, the data download and processing ... as well as the tools / software used.
What are the advantages of using NTL data and not for example Landsat 8 or Sentinel 2?
In general, considering the objectives, it seems to me a complicated process.
Author Response
Response to Reviewer 2 Comments
Point 1: 2.2.2 POI DATA
Baidu map? maybe is https://en.wikipedia.org/wiki/Baidu_Maps ?
It is better to indicate what it is.
Response 1: Thank you for your comments. It is Baidu Map Service. We corrected it as follow:
The POI data used by the research institute were derived from Baidu Map Services3, the most used and largest web map service provider in China [24]. We extracted POIs in May 2018 with the help of the application programming interfaces (APIs) that were provided by Baidu Map, concluding 343,064 record in total.
Point 2: Can the POI information be downloaded? What process have you adopted?
Response 2: Thank you for your comments. We downloaded the POIs with the help of the application programming interfaces (APIs) that were provided by Baidu Map. The POI data concludes the fields of name, detail address, longitude, latitude and category (hospital, education, amusement, restaurant etc.). As 3.1.2 said, we have adopted the process of deduplication, correction, and field investigation verification, We then imported the POIs into ArcGIS according to longitude and latitude information and were transformed to the WGS84 projection
Point 3: Also how many POIs are in May 2018 on Baidu?
Response 3: Thank you for your comments. There are 343,064 POIs in total (mentioned in 3.1.2). We added this in the first paragraph in 2.2.2.
Point 4: And what is Taile map downloader?
Response 4: Thank you for your comments and we are sorry for confused you. Taile map downloader is called ArcTiler (http://www.arctiler.com/p_downloader.html). It is a map download application. We corrected the expression of its name in the paper.
Point 5: This part must be well explained to understand the data source, the completeness, the data download and processing ... as well as the tools / software used.
Response 5: Thank you for your suggestions. The data preprocessing method is in section 3.1 at present. We are afraid that if this part was moved to section 2, there might be some confusion because we used the auxiliary data during the NTL preprocess while the introduce of the auxiliary data was behind the introduce of NTL data.
Point 6: What are the advantages of using NTL data and not for example Landsat 8 or Sentinel 2? In general, considering the objectives, it seems to me a complicated process.
Response 6: Thank you for your comments. First, the nighttime light (NTL) data enabled more detailed inner-city structure monitoring [1] than the multispectral images. And then for the study of urban structures, nighttime light images have been considered as a new potential source [2]. We think the NTL data can reflected the location of human activities better.
[1] Cai, J.; Huang, B.; Song, Y. Using multi-source geospatial big data to identify the structure of polycentric cities. Remote Sens. Environ. 2017, 202, 210-221 .[doi10.1016/j.rse.2017.06.039]
[2] Elvidge, C.D.; Baugh, K.E.; Zhizhin, M; Hsu, F.C. Why VIIRS data are superior to DMSP for mapping nighttime lights. Proc. Asia-Pacific Adv. Netw. 2013, 35, 62. [doi10.7125/apan.35.7]
Author Response File: 
Author Response.pdf
Reviewer 3 Report
The authors have used the VIIRS nighttime lights and Point of Interest data to detect urban centers and sub-centers of Hangzhou, China. I have suggested some edits and have some comments in the attached pdf. My main concern is the authors haven't specified whether they have used monthly or annual VIIRS data. Thanks.
Comments for author File: Comments.pdf
Author Response
Response 1: Thank you for your comments. We revised the manuscript according to your comments in the attached pdf. Here are illustrations about the VIIRS data and Google satellite image:
- We are sorry for confused you. The VIIRS data we used is monthly average data in May 2018. It is one composite image that made from several daily images in a month.
We described this in 2.2.1 as:The nighttime light (NTL) data used in our study were the NPP-VIIRS monthly average radiance composite images from the VIIRS Day/Night Band (DNB), acquired in May 2018.
And we added more details in 2.2.1 as: It was one composite image that made from several daily images in a month.
The relative expression in 3.1.1 changed as: the monthly average composite nighttime lighting image.
- One of the auxiliary data we used is Google satellite image. We revised the expression. It is the satellite image in Google Map. The satellite images provided by Google Map are basically the same as those provided by Google Earth.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Dear Authors,
Thanks for the detailed answers to the questions. Most of my questions have been well addressed. Reading again the paper, I would have some final issues to be improved, which relate mostly to language and adding some recent publications in the field.
Abstract: the first two sentences do not read well and would require language revision/shorting – For example you could change to: “The worldwide development of multi-center structures in large cities is a prevailing development trend. In recent years, China's large cities developed from a predominantly mono-centric to a multi-center urban space structure.”
Line 48 etc: also access to spatially disaggregated (statical) data might be a problem.
Paragraph 36-45: Please define here (1-2 sentences) what is the definition of centre in China.
Paragraph 52-60 misses several recent references:
https://www.mdpi.com/2072-4292/11/3/310
https://www.tandfonline.com/doi/full/10.1080/22797254.2019.1617642
Line 119-120 the newly added sentence has grammar issues – please correct. For example: “It was
a composite image made from several daily images in a month.”
Line 140 (and at many other places): these are Google Earth images – Google Earth is accessing images, e.g., from Digital Globe. The common way to refer to these images is “Google Earth images”.
Line 145: reads vague - which data? The data mentioned above? or more data?
Line 147: In principle, this violates privacy! Thus it is very good that you aggregated the data, could you please explain how did you aggregated them to 500 m grids?
Line 154, 243 and several places – please refer in a consistent way to the Local Moran I / Anselin Local Moran's I (LMI)
Line 162: This reads also a bit vague – I guess you mean you extracted the night time light images of your study area using the global VIIRS image and the administrative boundary of your study area?
Line 206: This question has not been answered by the revision – why ESP was not used: https://www.tandfonline.com/doi/full/10.1080/13658810903174803
Line 229: please refer to “equation” (not formula) and please define the variables used in equation 1-3.
Line 409: I still (also after reading your reply) do not understand whether the airport area has residential population?
Author Response
Point 1: Abstract: the first two sentences do not read well and would require language revision/shorting – For example you could change to: “The worldwide development of multi-center structures in large cities is a prevailing development trend. In recent years, China's large cities developed from a predominantly mono-centric to a multi-center urban space structure.”
Response 1: Thank you for your comments. We revised it.
Point 2: Line 48 etc: also access to spatially disaggregated (statical) data might be a problem.
Response 2: Thank you for your suggestion. We added it.
Point 3: Paragraph 36-45: Please define here (1-2 sentences) what is the definition of centre in China.
Response 3: Thank you for your comments. We revised it as follow: Urban center is a large and densely populated urban area and may include several independent administrative districts [1]. Definitions of urban center. Available online: https://www.vocabulary.com/dictionary/urban%20center (accessed on 30th Jul 2019).
Point 4: Paragraph 52-60 misses several recent references: https://www.mdpi.com/2072-4292/11/3/310 https://www.tandfonline.com/doi/full/10.1080/22797254.2019.1617642
Response 4: Thank you for your comments. We revised it.
Point 5: Line 119-120 the newly added sentence has grammar issues – please correct. For example: “It was a composite image made from several daily images in a month.”
Response 5: Thank you for your comments. We corrected it.
Point 6: Line 140 (and at many other places): these are Google Earth images – Google Earth is accessing images, e.g., from Digital Globe. The common way to refer to these images is “Google Earth images”.
Response 6: Thank you for your suggestion. We corrected it.
Point 7: Line 145: reads vague - which data? The data mentioned above? or more data?
Response 7: Thank you for your suggestion. We revised it as follow: The Google Earth image downloaded need no further processing.
Point 8: Line 147: In principle, this violates privacy! Thus it is very good that you aggregated the data, could you please explain how did you aggregated them to 500 m grids?
Response 8: Thank you for your comments. This data is produced by governmental agency and we are authorized to use it only for study and research. We made a gridded map by using the fish net tool in ArcGIS and counted the number of population points in each grid.
Point 9: Line 154, 243 and several places – please refer in a consistent way to the Local Moran I / Anselin Local Moran's I (LMI)
Response 9: Thank you for your suggestion. We revised it.
Point 10: Line 162: This reads also a bit vague – I guess you mean you extracted the night time light images of your study area using the global VIIRS image and the administrative boundary of your study area?
Response 10: Thank you for your comments. We revised it as follow: The nighttime light image of the study area was extracted using the global VIIRS image and the administrative boundary of study area.
Point 11: Line 206: This question has not been answered by the revision – why ESP was not used: https://www.tandfonline.com/doi/full/10.1080/13658810903174803
Response 11: Thank you for your comments. The ESP is not included in the essential eCognition software we bought, so we are sorry that we didn't know about the ESP tool before. The ESP tool calculates roc-lv (rates of change of LV) of Local variance of image object homogeneity under different segmentation scale parameters to indicate the best segmentation parameter. And the shape factor and compactness factor needed in the ESP algorithm panel need to be known by individuals in advance, which can be roughly determined by experiments. Our idea to decided factors is similar to the one if we use ESP tool. We would consider to use ESP tool in further study.
Point 12: Line 229: please refer to “equation” (not formula) and please define the variables used in equation 1-3.
Response 12: Thank you for your comments. We revised it.
Point 13: Line 409: I still (also after reading your reply) do not understand whether the airport area has residential population?
Response 13: Sorry for confused you. We revised it. Our original intention was to explain the reason of why the centers in b1 (figure 11) are not pratical (line 408) from the population aspect. So we did not emphasize whether they are residential population. The reality is the airport has a large quantity of travellers and that there are some residential population near the Xiaoshan Airport (some communities). The population in the airport was not counted in the statistic data.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
This manuscript is about the application of the local Moran’s I index, geographically weighted regression model, Nighttime Light Data and POI Big Data to detect urban centers of Hangzhou. Below are my comments about this article:
Line 97-102, " The main objectives of this study were: ......(2) to use different data sources and threshold methods for comparative experiments to verify the accuracy of our method. (3) to compare the performance of our method with the evaluation Report of the master plan." However, the authors do not explain how to verify the "accuracy" of the method and do not compare the "performance" of the method in the article. The authors should revise the manuscript to enhance the statement. For example, the authors should use some indicator (ex. population, landuse data) to demonstrate that the detected urban center results are correct.
Line127-135, section "2.2.2 POI Data". Generally speaking, the point-of-interest (POI) is a point data which maybe belong different category (residential communities, school, park, bus station, oil station, ....) with different area scale and usage rate, these characteristics were related to the boundary of urban centers. In the article, the authors just count the number and calculate the density of POI is not suitable.
Line 136-138, section "2.2.3 Auxiliary Data", the statement: "A Google satellite map at Level 17 was used as auxiliary data were acquired by Taile map downloader, with a spatial resolution of 1.19 m." is unclear. Please explain why the Google satellite map is suitable for this study and the process of data.
Line 150-156, section " 3.1.1. Preprocessing of NTL Data", the authors should add a figure to present the overlapping result of remote sensing image of night lighting, Google remote sensing map and the administrative boundary vector data.
Line 170-171, "...we conducted the multi-resolution segmentation through the multi-resolution models in eCognition, which is the first object-based image analysis commercial software.". Is it possible for us to conduct the multi-resolution segmentation without " eCognition "? How to do?
Line 178-179, what do scale factor, shape factor, and compactness factor mean? How to calculate? Please explain them in detail.
For the analysis of local Moran’s I, the authors should present how to create the weighted distance matrix and add a map of the result of local spatial autocorrelation analysis with the spatial unit and the statistics of each cluster segmentation
Line 287-288, " Figure 4. NPP-VIIRS nighttime light intensity map of Hangzhou in May 2018. ", the authors should explain the spatial pattern and statistics of NPP- VIIRS NTL Intensity value.
Author Response
Response to Reviewer 1 Comments
Point 1: Line 97-102, " The main objectives of this study were: ......(2) to use different data sources and threshold methods for comparative experiments to verify the accuracy of our method. (3) to compare the performance of our method with the evaluation Report of the master plan." However, the authors do not explain how to verify the "accuracy" of the method and do not compare the "performance" of the method in the article. The authors should revise the manuscript to enhance the statement. For example, the authors should use some indicator (ex. population, landuse data) to demonstrate that the detected urban center results are correct.
Response 1: Thank you for your comment. We added the accuracy of comparison experiment in 4.4 as follows: Our method detected two main centers and ten subcenters, while there were only two subcenters, located in Linping and Yuhang distinct, that had been detected in b2.
Besides, we demonstrated the effectiveness and accuracy of our result in Part 5(Discussion). By calculated the number of corresponding centers, we compared the experimental results with the urban center system proposed in the current Hangzhou City Master Plan Evaluation Report. This report was made by the government and the city centers it refered were calculated through population data statisticed in the end of 2017. So the city centers in the evaluation report could be seen as the reference centers.
Point 2: Line127-135, section "2.2.2 POI Data". Generally speaking, the point-of-interest (POI) is a point data which maybe belong different category (residential communities, school, park, bus station, oil station, ....) with different area scale and usage rate, these characteristics were related to the boundary of urban centers. In the article, the authors just count the number and calculate the density of POI is not suitable.
Response 2: Thank you for your suggestion. We agree with your advise. There are several characteristics of POI that are related to the boundary of urban centers, besides the number and density, for example, the scale of area, the usage rate, the relationship with people activities, the used ages and so on. However, in this study, we aimed to focus on the spatial pattern of POI, so we just studied about the number and density. We added your advise as the limitation of the study in Discussion as follows: There are several characteristics of POI that are related to the boundary of urban centers, besides the number and density, for example, the scale of area, the usage rate, the relationship with people activities, the used ages and so on. In this study, the spatial pattern of POI was the focus. Therefore, other charcteristics of POI data might be used to measure more precise boundaries of city centers combine with nighttime light data in the future.
Point 3: Line 136-138, section "2.2.3 Auxiliary Data", the statement: "A Google satellite map at Level 17 was used as auxiliary data were acquired by Taile map downloader, with a spatial resolution of 1.19 m." is unclear. Please explain why the Google satellite map is suitable for this study and the process of data.
Response 3: Thank you for your comment. We used the Google satellite map to be a reference during the preprocess of NTL data. Taile map downloader is a paid application and it has several remote sensing data and products for users to download. Each type of the data has different resolution levels. Level 17 Google satellite map has a spatial resolution of 1.19 m. It was precise enough for our study as a reference map and easy to get. As for the process of data, we input the boudary shape file of study area to the Taile map downloader and chose the projection we need before downloading. So the result of downloading could be used directly. We added the explaination in 2.2.3 as follows: Taile map downloader provided several types of remote sensing data and products. Each type of the data had different resolution levels for users to choose. We input the boudary shape file to the downloader and chose WGS84 projection. The result of downloading need no more process.
Point 4: Line 150-156, section " 3.1.1. Preprocessing of NTL Data", the authors should add a figure to present the overlapping result of remote sensing image of night lighting, Google remote sensing map and the administrative boundary vector data.
Response 4: Thank you for your suggestion. We rivised 3.1.1(Preprocessing of NTL Data) and 4.1(NTL Data Preprocessing Result). But we are sorry that we considered that the overlapping result was a mid-result in preprocess and the figure you suggested may be similar with Figure 4.
Point 5: Line 170-171, "...we conducted the multi-resolution segmentation through the multi-resolution models in eCognition, which is the first object-based image analysis commercial software.". Is it possible for us to conduct the multi-resolution segmentation without " eCognition "? How to do?
Response 5: Thank you for your comment. In this study, we need the segmentation result for the following statistics work. And the eCognition is a professional software about segmentation and classification. So we conduct the segmentation with eCognition. We think we can study on the comparision of segmentation algorithms in different softwares in the futher work.
Point 6: Line 178-179, what do scale factor, shape factor, and compactness factor mean? How to calculate? Please explain them in detail.
Response 6: I am sorry to confuse you and thank you for your comment. We explain the mean of scale factor in the first paragraph in 3.2.1, and the mean and calculate method of the shape factor and compactness factor in the first paragraph in 3.2.2.
Point 7: For the analysis of local Moran’s I, the authors should present how to create the weighted distance matrix and add a map of the result of local spatial autocorrelation analysis with the spatial unit and the statistics of each cluster segmentation
Response 7: Thank you for your suggestion. We revised 4.3.1 part and added a figure as follows: In the study, the weight matrix did not be considered. The result of local spatial autocorrelation analysis with the spatial unit and the statistics of each cluster segmentation was shown in Figure 9.（The figure can be seen in word file）
Point 8: Line 287-288, " Figure 4. NPP-VIIRS nighttime light intensity map of Hangzhou in May 2018. ", the authors should explain the spatial pattern and statistics of NPP- VIIRS NTL Intensity value.
Response 8: Thank you for your suggestion. We revised this part as follows: The NPP-VIIRS NTL intensity value ranged from 5.51 to 302.99. It was found that, the red pixels with the highest NTL intensity were located in Xiaoshan airport. And the pixels in the central part area (near West Lake) had higher NTL intensity than those in the marginal area. Besides, the most part of high intensity pixels in yellow were concentrated in the northeast part to West lake and to the banks of the Qiantang River, while the low intensity pixels in blue were mostly distributed along the border, especially in the northwest, east, and southwest part of the urban area.
Author Response File: Author Response.pdf
Reviewer 2 Report
Dear authors,
please revise the following:
Lines:
60-68 analyzes the existing literature on various socioeconomic aspects (urbanism, poverty, energy use) and introduces only two references regarding environmental issues and they are related to fishing (10,11); or the authors eliminate those two references or expand to more fields - i.e. Qingxu Huang et al. en "Application of DMSP/OLS Nighttime Light Images: A Meta-Analysis and a Systematic Literature Review, Remote Sens. 2014, 6, 6844-6866; where it talks about uses in fishing, wildfires, damages by natural disasters, or even luminescence of bacteria in oceans.
105-113 confusing; what is each region, area and urban? or give a further explanation in the text or re-arrange Figure 1.
124 says it has a pixel resolution of 15 ", this should be wrong since that is about 38 cm which seems exaggerated. Please revise this data.
137-138 indicates that it uses the Google satellite; however, also says the resolution is 1.19 m. Thus, if is correct please provide a reference.
The section "5. Discussion" appears without any reference; please add some.
best,
Author Response
Response to Reviewer 2 Comments
Point 1: 60-68 analyzes the existing literature on various socioeconomic aspects (urbanism, poverty, energy use) and introduces only two references regarding environmental issues and they are related to fishing (10,11); or the authors eliminate those two references or expand to more fields - i.e. Qingxu Huang et al. en "Application of DMSP/OLS Nighttime Light Images: A Meta-Analysis and a Systematic Literature Review, Remote Sens. 2014, 6, 6844-6866; where it talks about uses in fishing, wildfires, damages by natural disasters, or even luminescence of bacteria in oceans.
Response 1: Thank you for your comment. In the paragraph in 60-68 lines, we introduced 14 references on the study aspects of nighttime light data (reference[2] to [15]). In line 65, we began to analyze the topic of NTL in urban structure study. We considered that we did not “introduces only two references regarding environmental issues and they are related to fishing (10,11)”. So we are afraid that we might misunderstand your comment.
Point 2: 105-113 confusing; what is each region, area and urban? or give a further explanation in the text or re-arrange Figure 1.
Response 2: I am sorry to confuse you and thank you for your suggestion. We revised the expression in this part. We changed “Hangzhou, the capital of Zhejiang province, is one of the most prosperous regions” to “Hangzhou, the capital of Zhejiang province, is one of the most prosperous citys”, “administrative regions” to “administrative districts” and deleted the word “area”
Point 3: 124 says it has a pixel resolution of 15 ", this should be wrong since that is about 38 cm which seems exaggerated. Please revise this data.
Response 3: Thank you for your comment. 15 " meant 15 arc seconds in the article instead of 15 inches. We added a note with brackets in 2.2.1 as follows: 15" (arc second).
Point 4: 137-138 indicates that it uses the Google satellite; however, also says the resolution is 1.19 m. Thus, if is correct please provide a reference.
Response 4: Thank you for your comment. We downloaded the Google map by Taile map downloader. Taile map downloader provides different resolution levels of map for users to choose (the following figure can be seen in the word file). We considered the file size and the efficiency of map loading and then chose the Level 17 map with the resolution of 1.19m.
Point 5: The section "5. Discussion" appears without any reference; please add some.
Response 5: Thank you for your suggestion. We added two refernces(38,39) in the second paragraph in section 5.
Author Response File: Author Response.pdf
Reviewer 3 Report
It is a really good paper which is interesting to read.
As I found no significant drawbacks, I would suggest to accept the paper for publication with minor revisions.
Two things would help improving the paper:
a thorough language editing
The logic of the empirical approach is quite briefly explained. It would help the reader to get a few sentences on WHY the steps have been undertaken and what the exact AIMS are and how the steps are INTERLINKED. In many cases it is just an explanation of methodogical questions rather than logical reasoning.
Author Response
Response to Reviewer 3 Comments
Point 1: a thorough language editing
Response 1: Thank you for your comment. We used an English editing service provided by MDPI before last submitting and we chose another company’s editing service this time. However, there were something wrong during the payment process and it cost us about two days to email the editors and at last finished the payment. The polishing manuscript has not been returned so far.
Point 2: The logic of the empirical approach is quite briefly explained. It would help the reader to get a few sentences on WHY the steps have been undertaken and what the exact AIMS are and how the steps are INTERLINKED. In many cases it is just an explanation of methodogical questions rather than logical reasoning.
Response 2: Thank you for your suggestion. We added the reason of segmentation in 3.2 as follows: Although there were some administrative divisions, human activities were not limited by the administrative boundaries. So the new statistical units were needed. We revised the first paragraph of 3.3.1 as follows: The main center of the city was characterized by clustered plots with high population and activity densities, so based on the segmentation result of nighttime light imagery, a spatial clustering analyst method, the local Moran’s I, was calculated with the POI density of each unit. We added the link between the main center identification and the subcenters identification in 3.3.2 as follows: Therefore, there was a relationship between the human activity density of subcenters and the distance from the subcenter to the main center.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
This revised version has improved the manuscript clearly.
There are still two methodological issues as follows:
1. Line 100-101, "(2) to use different data sources and threshold methods for comparative experiments to verify the accuracy of our method." The authors still do not explain how to verify the "accuracy" of the method. In general, the accuracy should be a quantified value.
2. According to line 238-243, we can find that the spatial weight matrix is an important part to calculate the local Moran's I statistic. However, according to line 356, "In the study, the weight matrix did not be considered."...The authors should explain how to calculate the local Moran's I statistic without the spatial weight matrix.
Author Response
Response to Reviewer 1 Comments
Point 1: Line 100-101, "(2) to use different data sources and threshold methods for comparative experiments to verify the accuracy of our method." The authors still do not explain how to verify the "accuracy" of the method. In general, the accuracy should be a quantified value.
Response 1: Thank you for your suggestion. We added quantitative evaluation in part 5 as follow:
5.1. Delineation Accuracy
To quantitatively analyse the delineation accuracy of our method and comparison experiments, we evaluated the agreement between the coverage of the detected centers and the coverage of population census data. The population census data presents locations with human living. It could not stand for the spatial pattern of human activities for some extent.
The population census point data were producted in 2017. They were plotted by the coordinates onto a gridded map with the WGS84 projection and 462 m resolution, for consistency with the NTL data. Population records were calculated in each grid. We then used hot spot analysis and refered the hot spots, statistically significant spatial clusters of high values, as the referenced center coverage.
We conducted the confusion matrixs in ENVI and used overall accuracy and kappa coefficient yto measure the distribution agreement between the city center coverage from population data and the center results.
The results were shown in Table 1.(See table in the revised manuscript)
Point 2: According to line 238-243, we can find that the spatial weight matrix is an important part to calculate the local Moran's I statistic. However, according to line 356, "In the study, the weight matrix did not be considered."...The authors should explain how to calculate the local Moran's I statistic without the spatial weight matrix.
Response 2: Thank you for your comment and we are sorry for confused you. We wrote ambiguous sentences in the last revision. We meant that we did not input a weights matrix file in ArcGIS. We determind the weight matrix by inverse-distance method and then we did the standardization. The toolbox of local Moran’s I in ArcGIS is shown as the figure below. We revised this part in the manuscript as: In the study, the weight matrix was determined by inverse-distance method. It means that nearby neighboring features have a larger influence on the computations for a target feature than features that are far away. And then a row standardization was conducted. It means that spatial weights were standardized and each weight was divided by its row sum.
Author Response File: Author Response.pdf
