Accuracy of the Copernicus High-Resolution Layer Imperviousness Density (HRL IMD) Assessed by Point Sampling within Pixels

Round 1

Reviewer 1 Report

This study aims to assess the accuracy of the Copernicus High Resolution Layer Imperviousness Density by using a stratified random sample of ground truth points collected for Norway through very high resolution orthophoto. While the aim and methodological approach of the study are of scientific interest, the manuscript currently resembles a report rather than a scientific paper, and there is almost no analysis and discussion on the implications of the results. In this regard, the authors will have to invest in editing and rewriting several parts of the article. Below I provide some major comments and several specific comments hoping that these help to improve the manuscript.

 

MAJOR COMMENTS

-If you aimed to test for three specific hypotheses, ideally the methods and manuscript structure should be framed to tests and report the results of these hypothesis. Nevertheless, the manuscript suffers from several messy sections which seems to be disconnected, out of place or provide unnecessary information. This need to be revised to make the manuscript more cohesive and coherent.

-I think the most interesting part of the manuscript is the use of a within-pixel sampling strategy to estimate the percentage impervious per pixel. The author should place more focus on this method as this is, in my opinion, one of the strength and novelties of the study.

-There is almost not discussion on the discussion sections. Basically, the authors dedicate the vast portion of this section to speculate regarding what elements could be related to the under- or over-estimation of impervious surface, with no supporting evidence for the speculations. This section is descriptive and to narrowly focus on the specific results for Norway, with no relation to other studies or implications for using this or other classification products elsewhere. The study needs to be attractive and informative for a larger audience that goes beyond Norway. I do not think the paper achieve this in its current version.

 

SPECIFIC COMMENTS

-Line 26: Described or known as soil sealing?

-Line 29: “surface water, among other effects”. It would be also recommendable to include references for these effects. Perhaps a review on imperviousness effects may do the job.

-Line 29: as an example or as a driver?

-Lines 36-40: Provide the approximate dates of these mapping developments.

-Lines 42: The study…which study? I know you refer to the reference 11, but try to integrate these sentences to make them more fluid.

-Line 43: What was found by this review? Currently, that sentence is completely out of context and does not provide background information.

-Line 53: What is Pan-European? Not all readers will know what this mean.

-Line 59: Already mention the range in line 54.

-Line 69-70: This suggest that the accuracy of the product has not been assessed? Please clarify this in the text. If it was previously analyzed, report the findings.

-Line 75: Why using Norway and not another region? Europe has contrasting climates and landscape composition between regions, therefore analyzing the product in a single region may bias the results. Please provide explanation on this choice.

-Line 87: What do you mean by “reference material”? Reference for what in specific?

-Lines 93-94: The estimated imperviousness of the HRL IMD raster for Norway? Please clarify.

-Line 97: Is there any justification for using these strata instead of other?

-Lines 99-100: Same as previous comment, why selecting these number of points? I assume that is to make easier to compute percentages but is not detailed in the text.

-Table 1: Hectares are estimated from the HRL IMD layer? Also, to make the table easier to read, I suggest inverting the order of the strata, placing 100% on the top, and 0% on the bottom. This will avoid placing the 0% below the subtotal, which looks odd.

-Line 110: Should be QGIS

-Line 115: Was found or calculated/estimated?

-Line 119: What population? Also, start this paragraph by describing the objective of the following procedure. Something like “To estimate….a simple random sample was collected by…”

-Line 136-138: Provide the equations in separated lines.

-Line 141-145: What is important here, I guess, is if your data for each stratum has a normal distribution and your errors are not biased.

-Lines 147-158: I think part of this paragraph will fit better on the discussion section.

-Lines 161-162: Ok, but calculating the area of Norway is not part of the objectives of the study. This may be removed with no harm to the study.

-Line 166: Are you meaning multiplying by the proportion impervious? Then, should be “Weighted”.

-Line 168: This information is from Table 2? Table 1 show another percentage. Please check. These tables should provide the same information for the HRL%, but they have differences. Why is that?

-Lines 170-174: Where is this data shown? Tables?

-Lines 175-179: Where is this data shown? Tables?

-Lines 187-189: How can you infer this if there are no maps or analysis made for this?

-Line 190: There is an error in the Table2, it shows 0.00% in HRL.

-Line 195-202: This entire paragraph seems to be out of place. Furthermore, a large part of this should be placed within the captions for Table 2.

-Line 203-209: Figure 2 reports the same information than Table 2, thus unnecessarily duplicating the information.

-Lines 213-219: I do not think your data are reporting the omission error or accuracy, but just an average difference between the model and the true data. Of course, these will be tightly related, specifically the omission, but are not the same. For example, what are the omission rate and accuracy for the 70-79% stratum? If you aim to report omission and accuracy, calculate and provide the appropriate parameters.

-Lines 220-231: These lines are not necessary as they are covered in previous paragraphs.

-Lines 233-236: Avoid repeating or providing the results in the discussion.

-Lines 237-243: Is this a good or bad result? Is the figure provided by Statistics Norway more reliable than your estimates? If you do not provide this information, then the paragraph points to nothing.

-Lines 255-258: If you had calculated the commission/omission rate you could have had this information.

-Lines 272: Based on what information are you stating this?

-Conclusions: This section reads more like a summary than a conclusion. The only paragraph that seems conclusive is the last one.

 

Author Response

Please find my account of the revisions and comment resolutions in the attached document.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Author, I had the chance to read and evaluate your effort which is a well structured and conducted research. My comments are provided below that represent my personal opinion and do not intend to dispute the scientific soundness of your research, but to exploit the full potential of your work in case you decide to follow them. After reading the methods section, and especially the sampling scheme that was followed, some comments have risen from my side. There are two parts that, in my opinion, need to be more clearly explained or reformulated: The strong majority of the pixels represent locations that are non-impervious. This seems to have affected though the number of samples you randomly selected for your analysis, and you ended to have one stratum with 1000 samples, 1 stratum with 82 samples and 11 strata with 100. I would suggest keeping equal quantities per stratum (i.e. n_i<=82). I carefully read the explanation provided at lines 123-125 but this is a part that I disagree and suggest meeting the equality of proportions. Furthermore, please, in case that the number of strata for such purpose is a number proposed for the first time, provide more information about the method it was selected, or provide the corresponding reference.   Figure 2 does not seem to provide any further information since Table 2 overlaps it. Furthermore, in case you decide to keep it, please consider to change the "0-9%" label of x-axis to "1-9%"   From Table 2 we observe (and it is clearly stated from the author) that imperviousness higher than 40% is overestimated in HRL IMD. Could you provide an explanation about this observation? Could a correction be made to the acquired imagery. Thank you for your effort and good luck on your publication.

Author Response

Please find my account of the revisions and comment resolutions in the attached document.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors have answered to all my comments and taken into consideration most of my suggestions. When they did not agree with me, they provided a good explanation to maintain their point, which is always a good way to ensure the quality of the research. The paper has improved substantially, and focusing on the methodology make it more interesting for a wider public. I only have a few minor suggestions. The first is to remove the "The" from the beginning of the title. An second, to further place in context the results in the discussion sections. For example, the classification errors you find with your methods, are similar to those found by other studies? Solving this small issues, I am fine with publishing the article.

Author Response

Please see the attachment

Author Response File: Author Response.docx