Selection of Lunar South Pole Landing Site Based on Constructing and Analyzing Fuzzy Cognitive Maps
Round 1
Reviewer 1 Report
This manuscript describes the selection of a landing site at the lunar south pole in the framework of the upcoming Chang’E-7 (CE-7) mission, by considering several factors that are of importance: it has to close to a PSR region, with sufficiently small slopes, low rock abundance (no obstacles), and prescribed illumination and thermal environment conditions. The authors clearly describe the different conditions, and they introduce the used data through many instructive figures. They then present their results and discuss possible flight paths for the candidate landing sites. This work will be useful in future lunar south pole mission planning.
I have several comments, and they mostly pertain to asking for more information from the authors. I recommend publication after minor revision.
I have to indicate that I am not familiar with applying fuzzy logic, so some of my comments/requests for more background information may stem from this. Overall, the authors do not make clear to me why/how fuzzy logic is needed. I understand the different criteria, and ordering them in importance. Does subsequent masking (applying masks of terrain one after the other, after data are evaluated for the given conditions) leave one with no candidates? If so, doesn’t this also indicate something? In other words, the iterative process is not entirely clear to me; considering time, which could change illumination conditions, is not varied to estimate the right landing time (presumably).
I will list my comments here indicated with line numbers. They are a mix of comments on the content, and textual comments.
Line 22: add a space before “we” in “indexwe” (and add a comma)
Line 46: I am not sure if “siting” is the right word. Maybe just state, “Several studies into landing sites” or something like that.
Line 62: Change “Cabus” to “Cabeus”.
Line 93: which Antarctic test data set is meant here? This is kind of out of the blue. Or maybe the authors use Antarctic here for the lunar south pole. If so, I would suggest to leave Antarctic for the Earth’s south pole, and just call it lunar south pole.
Line 104/105 is not a complete sentence.
Line 126/127: which LRO/LOLA products are used? There is an extensive DEM called SLDDEM2015, and recent updates (see papers by Barker et al.) that could be of very valuable use for this study.
Line 170: maybe make clear if R is the local radius? It is not clear (to me) if a DEM is taken into account.
Line 179: PSRs have a narrower definition, as they would need to be in shadow for a long time span (hence, permanent). The time span is not clear to me in the explanations.
Line 184: I understand the requirement about the slope in the landing spot. How about the slope of the route to the PSR? Are they actually accessible? Maybe this will be explained later.
Line 220: the reference should be equation (9), not (8), I think.
Lines 284-287, equation (15): maybe a little more explanation for those not directly versed in fuzzy logic jargon? For example, how is mu defined?
Line 290: Maybe ‘family” should be “familiar”?
Line 302: Define FCM. This is the first time it is used. It is in the abbreviations list, but for clarity it would be good to introduce, and explain.
Line 308: how is the weight matrix defined?
Line 318: there is no W in equations 18 and 19, unless the authors mean omega. I am not sure how this relates to fuzzy logic. I can understand that each site, or even pixel, is evaluated for Ci, and given a score, but that seems straightforward.
Line 325: I think the authors here mean Figure 12 instead of 11.
Line 330: see earlier comment about Antarctic.
Figure 12: a lot of research has been undertaken to determine PSRs. How does this result compare to those? What time span for determining a PSR has been used? Was it necessary to redetermine them? Couldn’t a map of existing/known PSRs be used? Is there a reason to redo the analysis? Without sufficient detail, it may be difficult to determine if this PSR map is accurate enough. Yet in line 355, in the slope section, the authors indicate to use illumination data from a different study (Mazarico et al.). This should be cleared up.
The caption of Figure 14 should indicate the source of the data.
Line 365 is confusing, as the authors say there are neither permanent illumination or shadow zones. But Figure 12 indicates PSRs.
Line 384: I am not sure I understand the unit of km for cool zones. Is this a distance to? In what direction? Also, it is not entirely clear if Figure 15 was generated by the authors (using all Diviner data?) or not (using ref. 43). Indicate this, also in the caption of Figure 15.
Line 387: change “Figur” to “Figure”.
Line 398: I do not quite understand the iterations and maximum iterations. Is it coincidence that the max is 8 and that after 8 iterations equilibrium is reached? How does one know this is the case if the maximum is 8? Figure 16 shows convergence earlier than 8 iterations, and this makes more sense. I suggest to rewrite this bit, maybe pointing forward to Figure 16.
Table 4: the entry “Intal” should be “Initial’.
I am not sure I understand how the iterative process works, and some more explanations here could be useful. In a simple way, why wouldn’t the subsequent application of map masks based on the five criteria work? Each criterium presumably produces a map of suitable areas. Such a mask can be multiplied, and areas that remain are suitable candidates. Scores for each pixel can then maybe also be determined. It is just not clear to me where (and how) fuzzy logic comes into play. This could very well be due to my unfamiliarity with the math behind it!
Lines 442-443: it is not clear to me that Figure 17 shows that Amundsen had a better score. I understand that the scores will depend on the chosen landing time: local illumination and such will be of great importance. But it cannot (and should not) depend on the resolution of the data set, which seems to be suggested here, if my reading is correct.
Author Response
Response to Reviewer 1 Comments
Point 1: I have to indicate that I am not familiar with applying fuzzy logic, so some of my comments/requests for more background information may stem from this. Overall, the authors do not make clear to me why/how fuzzy logic is needed. I understand the different criteria, and ordering them in importance. Does subsequent masking (applying masks of terrain one after the other, after data are evaluated for the given conditions) leave one with no candidates? If so, doesn’t this also indicate something? In other words, the iterative process is not entirely clear to me; considering time, which could change illumination conditions, is not varied to estimate the right landing time (presumably).
Response 1: Thank you for your suggestions. We also found some minor problems in the language. We modified it in line 18, 55.
Point 2: Format modification.
Line 22: add a space before “we” in “indexwe” (and add a comma).
Line 46: I am not sure if “siting” is the right word. Maybe just state, “Several studies into landing sites” or something like that.
Line 62: Change “Cabus” to “Cabeus”.
Line 104/105 is not a complete sentence.
Line 220: the reference should be equation (9), not (8), I think.
Line 302: Define FCM. This is the first time it is used. It is in the abbreviations list, but for clarity it would be good to introduce, and explain.
Line 308: how is the weight matrix defined?
Line 325: I think the authors here mean Figure 12 instead of 11.
Line 330: see earlier comment about Antarctic.
The caption of Figure 14 should indicate the source of the data.
Line 387: change “Figur” to “Figure”.
Table 4: the entry “Intal” should be “Initial’.
Response 2: Thank you very much for pointing out our mistake. We also found some errors. We have highlighted the modified part in the paper.
Point 3: Line 93: which Antarctic test data set is meant here? This is kind of out of the blue. Or maybe the authors use Antarctic here for the lunar south pole. If so, I would suggest to leave Antarctic for the Earth’s south pole, and just call it lunar south pole.
Response 3: Thank you for your suggestions. We have changed it to the lunar south pole area in Line 93.
Point 4: Line 126/127: which LRO/LOLA products are used? There is an extensive DEM called SLDEM2015, and recent updates (see papers by Barker et al.) that could be of very valuable use for this study.
Response 4: Your advice is very professional. Since this paper focuses on the South Pole of the moon (80°s-90 °S), the range of SLDEM is (0-60°), which is too small. At present, partial data of the South Pole of the moon are available from LOLA DEM (5m/px), which is also used in this paper, but the overall data of the South Pole are available from CE-2 DEM.
Point 5: Line 170: maybe make clear if R is the local radius? It is not clear (to me) if a DEM is taken into account.
Response 5: Thank you for your suggestions. R is the moon's mean radius. DEM was calculated and reflected in the results in sections 3.2 and 3.4.
Point 6: Line 184: I understand the requirement about the slope in the landing spot. How about the slope of the route to the PSR? Are they actually accessible? Maybe this will be explained later.
Response 6: Thank you for your question. We supplemented it on line 449-451. China intends to make an in-situ measurement in PSR as part of the CE-7 Mission. A lunar lander carrying a mini-flyer will arrive at the sun-illuminated region (SIR) to provide solar power.
Point 7: Lines 284-287, equation (15): maybe a little more explanation for those not directly versed in fuzzy logic jargon? For example, how is mu defined?
Response 7: Thank you for your question. where µA(x) and µB(y) denote the membership value x to the linguistic term A and the membership value y to the linguistic term B respectively. For example, C1 and C2, C1 and C3, and so on.
Point 8: Line 318: there is no W in equations 18 and 19, unless the authors mean omega. I am not sure how this relates to fuzzy logic. I can understand that each site, or even pixel, is evaluated for Ci, and given a score, but that seems straightforward.
Response 8: Thank you for your question. ωj is the weight of the j evaluation index. For each Ci, a stable weight will be obtained by FCM, which will be normalized to calculate the score.
Point 9: Figure 12: a lot of research has been undertaken to determine PSRs. How does this result compare to those? What time span for determining a PSR has been used? Was it necessary to redetermine them? Couldn’t a map of existing/known PSRs be used? Is there a reason to redo the analysis? Without sufficient detail, it may be difficult to determine if this PSR map is accurate enough. Yet in line 355, in the slope section, the authors indicate to use illumination data from a different study (Mazarico et al.). This should be cleared up.
Response 9: Thank you for your question. For the determination of PSRs, we used Illumination data from a Different Study (Mazarico et al.). It is annotated at Lines 325.
Point 10: Line 365 is confusing, as the authors say there are neither permanent illumination or shadow zones. But Figure 12 indicates PSRs.
Response 10: Thank you for your question. This is a mistake in our writing. We have corrected it and it is on Line 372.
Point 11: Line 384: I am not sure I understand the unit of km for cool zones. Is this a distance to? In what direction? Also, it is not entirely clear if Figure 15 was generated by the authors (using all Diviner data?) or not (using ref. 43). Indicate this, also in the caption of Figure 15.
Response 11:Thank you for your question. Figure 15. Analysis of the maximum temperature at the South Pole of the Moon [44]. Map of maximum temperature profiles: (a). south pole; (b) 90°W-90°E; (c) 0°W-180°W.
Point 12: Line 398: I do not quite understand the iterations and maximum iterations. Is it coincidence that the max is 8 and that after 8 iterations equilibrium is reached? How does one know this is the case if the maximum is 8? Figure 16 shows convergence earlier than 8 iterations, and this makes more sense. I suggest to rewrite this bit, maybe pointing forward to Figure 16.
Response 12: Thank you for your suggestions. This is a batch result set by us, which is a stable batch condition. We have revised the original text, and it is on line399-403.
Point 13: I am not sure I understand how the iterative process works, and some more explanations here could be useful. In a simple way, why wouldn’t the subsequent application of map masks based on the five criteria work? Each criterium presumably produces a map of suitable areas. Such a mask can be multiplied, and areas that remain are suitable candidates. Scores for each pixel can then maybe also be determined. It is just not clear to me where (and how) fuzzy logic comes into play. This could very well be due to my unfamiliarity with the math behind it!
Response 13: Thank you for your question. First, the iterative process is to seek the optimal weight, set the membership function, fuzzy rules and so on. Second,First, the iterative process is to seek the optimal weight, set the membership function, fuzzy rules and so on. Second,A quantitative mathematical evaluation model refers to the measurement of the quantity involved in site selection (such as slope, light, etc.) with specific mathematical variables, which is a quantitative process. Kosko proposes a fuzzy Cognitive Map (FCM) by integrating Zadeh's Fuzzy set theory and Axelrod's Cognitive Map theory. FCM is a dynamic reasoning tool applied to site selection problems with expert participation. The advantage is that it can combine different factors and mathematical variables to find optimal weight. Line 214-280 is the process explanation of the algorithm, which is rarely attempted in planetary science. Most scholars analyze specific data to obtain the region of interest without applying the algorithm. Finally, this method is only a preliminary selection of areas to provide reference value. If further analysis is needed, high-resolution data are needed for analysis
Point 14: Lines 442-443: it is not clear to me that Figure 17 shows that Amundsen had a better score. I understand that the scores will depend on the chosen landing time: local illumination and such will be of great importance. But it cannot (and should not) depend on the resolution of the data set, which seems to be suggested here, if my reading is correct.
Response 14: Thank you very much for your question. Your suggestion is correct. The site selection problem is closely related to the launch time of the mission, which is not determined by the data resolution in a strict sense. We corrected Lines 443-lines 447.
Author Response File: 
Author Response.docx
Reviewer 2 Report
I would like to thank the authors for the updated version of the manuscript that addresses my major concerns. Language editing is still required, and this is the reason why I recommend the paper after minor revision.
Author Response
Response to Reviewer 2 Comments
Point 1: I would like to thank the authors for the updated version of the manuscript that addresses my major concerns. Language editing is still required, and this is the reason why I recommend the paper after minor revision.
Response 1: Thank you for your suggestions. We also found some minor problems in the language. We have highlighted the modified part in the paper.
Reviewer 3 Report
Please see attached file
Comments for author File: 
Comments.pdf
Author Response
Response to Reviewer 3 Comments
Point 1: Please explain why not to set wj equal each other (1/5).
Response 1: Thank you for your suggestions. FCM is a dynamic reasoning tool applied to site selection problems with expert participation. The advantage is that it can combine different factors and mathematical variables to find optimal weight. Line 214-280 is the process explanation of the algorithm, Wj is also a stable weight obtained by the FCM algorithm, and its importance degree is not consistent.
Point 2: Please explain in more understanding way how the wj were derived. My
impression is that current explanation is not optimal.
Response 2: Thank you for your suggestions. In order to make the reader better understand, we modify the description in Section 3.3.
Point 3: Please pay attention in your method that not all PSRs demonstrate presence of
near surface ice (see LRO observations).
Response 3: Thank you for your suggestions. It is not all PSRs demonstrate presence of near surface ice. We have also made corrections to the article by referring to literature on LRO and other observations, in lines 335-lines 336.
Point 4: I am a little bit lost in parameter C4 (illumination). Is this factor higher (better) for
the landing site with a higher illumination? If yes, then why Figure 7 shows that
feasible values belong to the interval [0, 0.5]?
Response 4: Thank you for your suggestions. Thank you very much for asking this question. HF should be 0.7-1. There was a problem in the drawing, and we have corrected it
Point 5: Please check carefully over the manuscript and fix some mistakes and misprints.
For example,
line 72-73 : “the annual surface temperature of the landing site is greater than
110K”. Is greater or less than 110K?
line 414: “lunar surface slope (C2) were 0.741, 0.701”. 0.741 or 714?. Table 4
says that it should be 0.714.
Response 5: Thank you for your suggestions. The second was that the annual surface temperature of the landing site is less than 110K, has a moderate slope (< 10), and is closer than 1km to the PSR. Line 414: “lunar surface slope (C2) were 0.714.
Author Response File: Author Response.docx
Reviewer 4 Report
Detecting lunar polar ice is always appealing to the planetary fellows. This work presented a nice method to select potential landing regions near the PSR for the China Chang'E-7 mission. This version of the manuscript has been greatly improved by adding more details about the method. I only have several minor suggestions, please see the attached pdf.
Comments for author File: Comments.pdf
Author Response
Response to Reviewer 4 Comments
Point 1: Line 79, “our focus is” to “focus”
Figure 5 lower right corner label for theta should be (m+1,n+1),
Line 189: “The surface slope” I think this is more like an average slope variation within a
3x3 box or the “terrain gradient” you’ve mentioned rather than the surface slope. Please
clarify that in the corresponding text.
Line 206: “under both the lander and the rover” change to “at the landing area”
Line 220: “ d is set to 3 m.” change to (d=3 m)
Lines 256-257: “The flatter the topographic 256 rock is, the better it is.” Change to “The
falter the topography ”
Line 325: Figure 12
Figure 13, was this slope calculated using the method in section 2.3.2?
Figure 13-15, if these maps are not made using methods presented in this paper, need
to cite references in the figure caption.
Line 439, “former”, did you mean the first two craters? if so, change to “the first two
craters”
Lines 443-44: “Here the most suitable sites are mostly distributed in the marginal areas,
and problematic areas are eliminated.”
Response 1: Thank you for your suggestions. We also found some errors. We have highlighted the modified part in the paper.
Author Response File: Author Response.docx
