Quantification and Analysis of Impervious Surface Area in the Metropolitan Region of São Paulo, Brazil
, Rúbia Morato 1, Marcos Martins 1, Guilherme Mataveli 1, Pablo Nepomuceno 1 and Marcos Martines 2Round 1
Reviewer 1 Report
Some small writing errors need to be corrected, careful reading of the MS will pick these up. For example, a missing comma on Line 130
Two contrary statements, lines 56 - 58 - in the first sentence, authors state that "watersheds damaged by ISA are primarily concentrated in the USA, Europe, Japan, China and India. On the other hand, pristine watersheds having little or no ISA are concentrated at Northern latitudes". - Please note that both statements cannot be correct as USA, Europe, Japan, China and most of India are in northern latitudes.
Lines 230 - 231 - why did authors only use 50 samples for such a large area?
Figure 3f and lines 283 - 287 - what are the purposes of the symbology for the different locales. Why are these important?
Lines 297 - 298 and Figure 5 - are the authors stating that they manually drew the curves on the graphs? By what methodology did they identify the appropriate location for the curves? It appears it was a best guess?
Starting on Line 402, it appears the authors are attempting to equate the ISF values with the level of economic development - primary, secondary, tertiary. Where did authors get the information on lines 406 - 411 and in Figure 8? Authors state on 405 - 406, there is no such relationship, so how were these numbers and the figure created? Is there a specific source that states that 60% of Biritiba Mirim GDP is in the primary sector? It appears the authors just threw this information in, it is not discussed in the abstract.
Author Response
Response to Reviewer 1
Dear Reviewer, thank you for reading the work and for your valuable comments. The changes are highlighted in yellow in the text.
Point 1: Some small writing errors need to be corrected, careful reading of the MS will pick these up.
Response 1: We have performed a careful final reading of the manuscript correcting individual typos and missing commas.
Point 2: Two contrary statements, lines 56–58 – in the first sentence, authors state that “watersheds damaged by ISA are primarily concentrated in the USA, Europe, Japan, China and India. On the other hand, pristine watersheds having little or no ISA are concentrated at Northern latitudes.” – Please note that both statements cannot be correct as USA, Europe, Japan, China and most of India are in northern latitudes.
Response 2:
Thank you for this observation. We have modified the text by deleting the part that mentions the reference to “northern latitudes”. The text is now as follows: “On the other hand, pristine watersheds having little or no ISA are concentrated in central Asia, portions of Africa, the Amazon Basin and the southern regions of South America and the Arabian Peninsula.”
Point 3: Lines 230–231 – why did authors only use 50 samples for such a large area?
Response 3:
We assume that 50 locations distributed in a random, stratified manner are sufficient to represent the variability of soil impermeabilization in the Metropolitan Region of São Paulo (MRSP). This number of samples is commonly used as a reference to obtain a consistent statistical analysis of the results. This information has been inserted into the body of the main text (lines 45-246).
Point 4: Figure 3f and lines 283–287 – what are the purposes of the symbology for the different locales. Why are these important?
Response 4: The purpose of the symbology used is to highlight the variability of ISA in different use types. By the positions of the symbols, it is possible to observe, for example, that while industrial and low-income residential areas have less variability of ISF, areas such as Morumbi and USP have high values of amplitude of ISF as a result of the size of the buildings and the diversity of materials present. This information is described in the text (lines 249-300).
Point 5: Lines 297–298 and Figure 5 – are the authors stating that they manually drew the curves on the graphs? By what methodology did they identify the appropriate location for the curves? It appears it was the best guess?
Response 5: The curves representing cumulative distribution of ISF for each municipality of the MRSP were generated automatically according to the procedures described in topic 2.4. ISF model integration. The curves were grouped visually, paying particular attention to the shape of the curve, because we believe that the shape of the curve shows the pattern of spatial organization and functionality of each municipality within the context of the metropolitan region. Tests could have been performed using statistical grouping techniques, thus making the procedure more objective, but this would have been outside the scope of our proposal.
Point 6: Starting on Line 402, it appears the authors are attempting to equate the ISF values with the level of economic development – primary, secondary, tertiary. Where did authors get the information on lines 406–411 and in Figure 8? Authors state on 405–406, there is no such relationship, so how were these numbers and the figure created? Is there a specific source that states that 60% of Biritiba Mirim GDP is in the primary sector? It appears the authors just threw this information in, it is not discussed in the abstract.
Response 6: The information related to the contribution of the primary, secondary and tertiary sectors to the GDP of each municipality is provided by Emplasa (a state-owned metropolitan planning company). This data source has been inserted into the header of Figure 8 (line 434).
We opened the discussion relating ISA to GDP because in Topic 1 – Introduction – we commented on the relationship that exists between the intensification of economic activities and the increase of ISA. In the majority of studies, this relationship is addressed at a more general level of detail; our objective was therefore to check whether it is possible to observe this relationship at the municipal level. Figure 8 was constructed using the data from Emplasa and from the groups classified according to the pattern of distribution of ISF. As already mentioned, we concluded that, for the municipal scale, the GDP has little potential for explaining the variability of ISA due to the fact that the municipalities of the MRSP have their economies heavily concentrated in secondary and tertiary sectors (this information is briefly described in the Abstract, lines 41-43).
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The spectral mixture analysis has been widely used in quantifying impervious surface fractions. I didn't see any new contribution to the methodology. The authors just applied it directly to a different study area. Therefore, I don't think that's enough to be published in the Remote Sensing.
Author Response
Dear Reviewer, thank you for reading the work and for your critique. Unlike other works that focus on the accuracy of estimation of ISA, the analysis in this study focuses on the integration of ISA information on the pixel scale with the spatial variability present in the urban planning units while maintaining methodological rigor when generating these estimates. In addition, a multiscalar analysis is performed by comparing the distribution of ISA at different pixel resolutions and in different geographical planning units. We believe that this study reinforces the use of information related to the impermeabilization of soil in the territorial planning by providing elements for a better understanding of the relationship between type of use and impermeabilization.
Author Response File: Author Response.pdf
Reviewer 3 Report
The study aimed at quantifying impervious surface areas in the Metropolitan Region of São Paulo using landsat 8 imagery.
This study is well-designed and the conclusions can be supported by the results. However, some details of the methodology are not given.
1. The color scale in figure 1 shows ISA fraction values, but what is the data source of the ISA fraction values? What are spatial resolution and date? In the fast developing area, the ISA may change rapidly.
2. Line 186: the authors did not use the original normalization method, does this cause any difference in the results?
3. Please give more details about the endmembers, like why they were chosen, vegetation types of Green and non-green vegetation endmembers.
4 How did the authors choose the reference sites? Please mark them on the map if possible. Providing a confusion matrix will help the readers to understand the results better.
Author Response
Dear Reviewer, thank you for reading the work and for your valuable comments. The modifications were highlighted in green.
Point 1: The color scale in figure 1 shows ISA fraction values, but what is the data source of the ISA fraction values? What are spatial resolution and date? In the fast developing area, the ISA may change rapidly.
Response 1:
Figure 1 shows the color scale representing the ISA fractions. The estimate of ISA was generated using Landsat-8 OLI imagery acquired on 23 August 2015. The spatial resolution of the image is 30m. This information has been inserted into the header of Figure 1 (lines 149-150).
Point 2: Line 186: the authors did not use the original normalization method, does this cause any difference in the results?
Response 2: A small change from the original method really has no effect on the results of the endmember selection, nor on the modeling of mixing. However, normalization by summing the values of the bands prevents certain classes from having spectral signatures with values greater than 1. This information is highlighted in the text (lines 187-190).
Point 3: Please give more details about the endmembers, like why they were chosen, vegetation types of Green and non-green vegetation endmembers.
Response 3:
The endmembers Dark, Bright, GV and NPV were selected in the modeling of mixing because they represent the spectra that are representative of the composition of the mixture of the landscape analyzed at the moment of image acquisition. The endmembers Dark, Bright, and GV can be considered “stable” spectra, because they are part of the composition of the dominant landscape throughout the year in the Metropolitan Region of São Paulo. The NPV spectrum is more present in the autumn and winter months. Due to the images used having been acquired in the winter period, the presence of dry vegetation (grass) in the urban fringe and in rural areas is very common. Therefore, the insertion of the NPV endmember is important to represent the areas of dry grass and avoid these areas considered permeable being modeled as fractions of the spectrum of high albedo (Bright). This information is highlighted in the text in Topic 4 – Discussion (lines 351-365). As a way of better clarifying the selection criterion to the reader, we have added some text to the paragraph in line 215. “These endmembers represent spectra that are representative of the composition of the mixture of the landscape analyzed and were selected in the following places: in a clean-water area, on an industrial roof and over a large area of green and dry pasture, respectively.” (lines 215-216)
Point 4: How did the authors choose the reference sites? Please mark them on the map if possible. Providing a confusion matrix will help the readers to understand the results better.
Response 4:
The samples were collected in a stratified random manner with the aim of avoiding sampling bias and taking into consideration different patterns of land use and vegetation cover. This information is described in lines 243–245.
We have decided not to include the reference samples in the map (Figure 1) because they make the map somewhat polluted and contribute little to the understanding of the methodology.
As to the analysis of the accuracies and errors of the estimates, the confusion matrix was not used because we chose to use a continuous evaluation model that is done by adjusting R2 and an estimate of the residue (reference value – estimated value).
Author Response File: Author Response.pdf
