A Tailored Approach for the Global Gas Flaring Investigation by Means of Daytime Satellite Imagery

Round 1

Reviewer 1 Report

‘A tailored approach for the global gas flaring investigation by means of daytime satellite imagery’ presents a dataset of large gas flaring sites across the globe, based on daytime observations from the Landsat-8 Operational Land Imager. As the authors state, global oil and gas flaring activity is an important source of greenhouse gases, and addressing it a potential climate mitigation strategy. There is a need for more observation-based assessment of global flaring, of which persistence is an important factor, so I find the work to be very timely and likely to be of interest to the scientific community. There are, however, areas of the manuscript that I believe need to be addressed prior to publication.

Presentation: Throughout there are many single-sentence paragraphs which I find distracting to the presentation of the material. I am not sure if this was intentional or the result of some formatting error, but I would suggest revising the text to combine sentences into paragraphs to improve the flow and clarity of the text. More importantly, at times the text also lacks sufficient context and details to support the statements made. I highlight several of these areas in my specific comments below.

Validation/Accuracy: Efforts to validate the dataset are not discussed in much detail. The authors do mention visual inspections of the hotspot locations with visual imagery. Is this the main evaluation method? Accuracy numbers are provided for previous works, but again with little detail. Are accuracy numbers from these previous iterations of the method expected to be representative for the dataset presented in this paper? The abstract mentions 99% accuracy, but that is not explicitly shown in this manuscript.

Specific Comments:

·       The abstract should clarify that the current dataset is biased toward large, high temperature flares (particular wording up for debate). The authors do state later in the text that this is an incomplete dataset, but making this point earlier is needed to properly interpret much of the discussion around global onshore/offshore flaring distributions and persistence.  

·       Lines 32-33: I agree that direct ‘on-site’ GF measurements are expensive, but how are they ‘poor’? Are they are poor representation of global flaring because there are so few? Or are they measurements themselves of poor quality for some reason?

·       Lines 35-36: Is the thought that satellite observations are ‘low-cost’ because they are based on existing or multi-use space-based instruments, where the satellite is already a sunken cost?

·       Lines 39-42: I am confused by the statements ‘as well as their recognition from space an open challenge’ and ‘from the satellite perspective, the GF identification is favored by their position invariance and temporal-consistency.’ I would suggest taking a second look at the wording here for clarity.

·       Paragraph, lines 56-64 – Starting of a paragraph with ‘On the other side’ is a bit of awkward wording. As the reader, I was searching for what the ‘other side’ was in reference to.

·       Line 77: reliability of 88% compared to what?

·       Section 2.1 – What do the various tiers represent? With the inclusion of ‘Real Time’ one can assume is some sort of temporal designation, but no further details are included.

·       Line 113: Digital Number? Scaled radiance? These terms are used with little context for the reader to infer their meaning.

·       Eq.3: The formatting appears off and/or missing the flag values.

·       Line 161: ‘ed’ typo?

·       Line 162: ‘such kind of “punctual” industrial hotspots’ Without further details I do not know what this means.

·       Line 169: What time intervals are used to create the cumulative mask?

·       Line 177: ‘[with] respect to’

·       Line 179: “the one guaranteeing the better detection accuracy” This statement is vague and more details are needed to support this statement.

·       Figure 1: I am unclear about what you call a ‘site’ in the figure and the text. For flaring, is a ‘site’ a single flare or could it be multiple flares? Also regions of high flaring in the USA (Texas, North Dakota) do not appear to show up on Fig.1 as flaring hotspots which is surprising given the large number of flares and flaring intensity in those regions. Therefore, I do not really know what conclusions to take away from this figure. Perhaps that these are the areas with the largest/hottest flares?

·       Lines 204-206: Eq. (3) does not ‘suggest’ a thermal anomaly if one of the three conditions are met, it defines those conditions are requirements. I would recommend reconsidering the word choice here. Whether or not these conditional requirements accurately capture actual thermal anomalies is other question. 

·       Eq. (4): There appears to be another formatting issue here. This is first usage of the summation notation, and explanation of what this represents should be included in the text.  What does HP_EP stand for? ‘Extreme Pixel’ I am guessing, but I do not recall seeing that defined anywhere.

·       Table 1: How are the various conditions (a-f) assigned to the sectors (e.g. gas flares, industrial site, etc.)? Visual inspection of satellite imagery?

·       Line 224: I would be hesitant to use the word ‘perfect’ here, especially since Table 1 shows for condition ‘b’ 1% was assigned to industrial sites.

·       Occurrence Frequency & Persistence level: Are these normalized by how many high quality observations are made of the scene? For instance, what if two flaring sites have differing cloud coverage on the annual-scale, but otherwise the same flaring persistence? Will that show up in the ‘Occurrence Frequency’?

·       As mentioned in my first comment, the presentation of the global distributions of flaring (figures 4 & 5) needs to include a discussion of the incompleteness of the dataset, since flare volumes are not included in your numbers due to missing smaller flares, which when considered in aggregate may affect your distributions.   

·       Figure 5. ‘The most GF-affected countries by continent.’ There appears to be typo since this figure is not broken down by continent. Also, ‘GF-affected’ is a bit odd. Maybe instead you could say something along the lines of ‘Percentage of global gas flares detected by country.’

·       Lines 405-406: How has the DAFI flaring sites been ‘integrated’ into the VIIRS nighttime one? Is this in reference to the Figure 12 or have the two products actually been combined in some way?

·       Lines 410-411: ‘A general satisfying spatial agreement between daytime and nighttime GF maps catches the viewer’s attention.’ This is not a very rigorous statement. I think it is more fair to say that qualitatively there are spatial similarities between the datasets, such that general flaring regions identified by DAFI appear to also be captured by VIIRS/VNF.   

·       Line 412: “punctual comparison”? I do not know what this means.

·       Are VNF and VIIRS used interchangeably throughout? Everything is defined in the text, but I recommend being careful with swapping between the two acronyms for the sake of clarity.

·       Lines 451-455: ‘The new global daytime map of gas flares, which is expected to be more realistic than the currently developed, both in terms of accuracy and sensitivity, will be uploaded on the website.” This paragraph is unclear to me. Which is the new global daytime map, the data shown in this paper or a future dataset? What website?

Author Response

Please, see the attached file

Author Response File: Author Response.pdf

Reviewer 2 Report

Dear authors,

Please find attached my comments.

Comments for author File: Comments.pdf

Author Response

Please, see the attached file

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thank you to the authors for incorporating my feedback into the revised manuscript. The specific comments I raised concerning clarity have been addressed in the new version of the text. There are some minor grammatical issues remaining, but I assume those will be addressed in final proof reading of the article. Many single sentence paragraphs also remain, however, given the many clarifications/revisions made throughout the text, I find what appears to be a stylistic choice by the authors less distracting. Finally, the addition of the ‘DAFI website’ section significantly improves the readability and utility of the paper, and making the features illustrated in Figures 6-10, and correspondingly the whole dataset, much clearer.