Visual Perception Optimization of Residential Landscape Spaces in Cold Regions Using Virtual Reality and Machine Learning
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis research presents valuable insights by exploring residents' perceptions of recreational spaces through systematic measurements from eye tracking and VR environments, thereby circumventing the subjectivity inherent in traditional questionnaire surveys. However, the clarity and readability of the paper suffer due to its poor writing style and word choices. The article's organization and flow could be improved significantly. Below are some specific suggestions.
Line 31-33: “The health quality …” sentence could be rephrased for clarity and conciseness.
Line 49-50: The transition from discussing the benefits of leisure activities to focusing on light environment indicators feels abrupt.
Line 68-69: specify which methods are “traditional methods”
Line 111-112: consider breaking into two sentences.
Line 122: ANN is not explained
Line 143: Acronyms and abbreviations should be explained in the figure title or footnotes
Table 1: what is the greening rate? I assume it’s the green space ratio?
Line 167: give examples of material attributes.
Axis labels in all figures are too small to read.
Not all terms need to be abbreviated. It makes the article very crowded and difficult to read. For example, Line 504 to Line 519
Section 2.5 should be moved to the appendix or removed from the article, as it serves no substantial purpose. The author should focus on KNN and GA as it seems the main conclusions are drawn from the results of these approaches.
Insights into the demographics of the study participants (age, cultural background, socioeconomic status, etc.) and how these factors might influence visual perception and satisfaction would add depth to the analysis. This is crucial for understanding the applicability of the findings across different populations.
It is recommended that the author consider engaging professional language editing services to refine and copy-edit the manuscript before it is resubmitted.
Author Response
Thank you for recognizing our team's research efforts. Our team has revised and improved the manuscript according to your suggestions, please see the attached word file for the response.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsSummary
The purpose of this research is to better understand the difference in the visual experience of residential outdoor space with and without snow in a cold region. In itself this is a useful incremental contribution. However, the larger contribution of this study is the thoroughness of the research design. There are essentially four parts:
1. Characterize. Identify adjectives that best describe photographs of residential outdoor spaces. Have a large number of residents use sematic differential ratings to describe physical aspects of their residential outdoor spaces.
2. Experience. Investigate viewing behavior on a two-minute walk through four residential outdoor spaces in snow and non-snow seasons using eye-tracking (Tobii glasses) in the field.
3. Validate VR. Create 3D VR environments duplicating each of the four residential outdoor spaces. Respondents repeated the two-minute walk using HIVE headset.
4. Evaluate factors. Create 3D VR environments for two sites representing variation in the six most important residential outdoor space factors identified in parts 1 and 2, using an orthogonal design. Have participants virtually “walk” through it similar to how the actual spaces were experienced. Use eye-tracking to record the visual experience.
5. Analysis. Validate the VR experience against the in-field experience. Conduct an analysis using each of four machine learning approaches to explain visual perception satisfaction scores and compare the results.
The results indicate that the KNN approach, closely followed by the ANN approach best model visual perception of these spaces. Ranges of optimal values are given for six factors used to describe these residential outdoor spaces in two seasons. The interactions among these factors are not investigated. Such interactions are not just statistically important, but can provide important strategies for design. However, I think that the most important result is the comparison of on-site and VR eye-tracking as an approach to validate the VR experience.
I may have some of this wrong, for instance I am not sure I understand when respondents are on site in the field, looking at photos, or using VR. I found it difficult to always determine this, even after more than one reading (for instance, after the second reading I am not sure that photos are use at all). This suggests that the text needs to be clearer. The research design is complex. The consistent use of terms will help reduce confusion. Use headings in the text and Figure 1 to reflect the research design outline.
General Comments
1. This is an ambitious project, with several very important contributions! Thank you for your hard work.
2. There is still a need for additional editing and improvement in the use of English. One suggestion is to decide on the key terms and then use them consistently. For instance, it cold area or cold region. Or should it be two seasons or cold climate? It does not really matter so much which is used, just be consistent. Other suggestions are in the Specific Comments.
3. If part of the difficulty is that there are multiple ways to translate some Chinese words or phrases, it may be useful to selectively include the Chinese in parentheses. It would be a shame to loose meaning because of the need for translation. However, it is also possible that this suggestion is be helpful and more a sign of my ignorance.
4. The instructions to authors specifies that the minimum font size is 8 pt. While it may not be possible to do this for all text in the figures, please try to do as much as possible. There is a lot of helpful information in the figures, but it is difficult to read them.
5. The article seems very long. I suggest critically reviewing the Introduction to remove text that is not directly relevant. In other places there needs to be more explanation. For instance, the description of the four machine learning approaches is not sufficient. It may be better simply to cite the software and others articles that provide a clear detailed description of each one.
6. Make sure that abbreviations are all identified the first time the full text is used (excluding the abstract). I personally appreciate that you use both the full text and abbreviation when space allows. However, the journal’s style guide may disagree with me.
7. Figure 1 is very important. Make sure that the text in each box corresponds to the paper’s headings and text. Again, consistency in how terms are used will make it easier to follow what you are doing.
8. The Semantic Differential is typically a seven point bi-polar scale (i.e., -3, 0, +3) anchored by a pair of adjective antonyms; but it is also more than that. The traditional application would have a respondent use many of these adjective-pair ratings to describe something (e.g., a place, a meal, a person). The data were often analyzed through factor analysis; three factors dominated the results across multiple studies: evaluation, potency and activity (Osgood et al. 1957. The measurement of meaning). In this study you are using a bi-polar scale, but pre-judging the attribute’s meaning by only using one adjective pair. While not a big thing, the “Semantic Differential” implies a method that goes beyond using an adjective pair to anchor a bi-polar scale.
9. The study focuses on predicting the effect of several variables on leisure visual perception/satisfaction. From a design perspective it would be important to also consider the interactions among these variables.
10. The study seems to rely primarily on a two-minute walk through four residential outdoor sites during snow and non-snow seasons. It might be helpful to see a plan view of the sites with the starting and ending points. Also matched photos from the site and the 3D VR model.
11. It is difficult to clearly understand the number of observations in each part of the study. There are 4 outdoor spaces, and 2 seasons (i.e., snow, non-snow) in phases 1 through 3; phase 4 has 2 sites. (See the Summary at the beginning of this review for the phases.)
A different number of subjects are used in each of the three data collection phases. The first phase has responses from 663 residents. The second phase 50 participants used Tobii glasses, but it is unclear if they all viewed all four sites. The third phase has 32 participants actually walked the site (line 282) and used HIVE headsets to view the 3D VR model in both seasons. There are “8 scenes (i.e., 4 sites and 2 seasons?) and 8 people in each scene. The fourth phase is the Orthogonal Experiment that uses 2 sites in 2 seasons and simulates the variation in 6 factors represented by 49 “scenarios” in each season. There are 40 participants who used some sort of unspecified VR equipment.
12. How is the dependent variable (e.g., visual experience of residential outdoor space or visual perception score) measured in each phase? I assume it is question 5 in the questionnaire used for the residents in phase 1. However, there is no indication of how it is measured in the other phases.
13. I would suggest that in a spatial environment, there is no single “optimal” composition. This may be a relevant concept for a static view of a particular class of landscapes, but variation is also an important characteristic of a spatial composition. If every view represented the “optimal” condition, then the environment would be monotonous and unpleasant. This also suggests the importance of the interaction among factors in design composition.
14. I am interested in the use of “machine learning” in this way. Is it the new fashion because of the ascendence of AI? For instance, regression is a statistical method that does not require a “machine” and was developed before computers. It is normally applied to continuous data, and I would not consider it a “classification” process. When I think of machine learning, I think of conducting an analysis with a training data set and then validating it against a prediction dataset. This is not done in most of the research that uses regression, though perhaps it should be!
15. I suggest rewriting the 2.5.1 Machine Learning section. Make sure that each method has a citation that very clearly describes how it is calculated. Make sure that the assumptions, limitations and advantages of the method are stated. Perhaps give an example of where it has been applied in the landscape perception research literature. The given equations are not very helpful, since some of the variables are not defined.
16. Are the classification methods more “accurate” because they are less precise—they treat the independent data as 5 ordinal classes rather than a continuous measurement?
Specific Comments (by line number)
012 Suggest changing “lack of ignoring” to “ignoring”.
027 Citation [27] does not appear to have investigated perceptions in the physical landscape. Rather they compared VR and photographs.
038 Suggest replacing “proven” with “demonstrated”. It will take may replications before it can be considered proven.
045 Is the affect always a positive influence on visual satisfaction or is there an optimal height, for example? Perhaps just delete “positive”?
046 The point of [10] is that the research suggesting vision as the universally dominant perception may be due to available technologies to study vision, and socio-cultural biases. This is an interesting argument, but does not seem to support your choice of visual perception for investigation.
046 Insert [11] at the end of the sentence to cite the Wen et al. article.
053 The citation [17] is an incredibly important line of research as we work to make VR experiences as much like in-field experiences as possible. A next step beyond eye-tracking would be investigating brain-wave patterns.
064 Should there be citations of the relationship between landscape and health? Maybe a review study?
070 This may be true [23], but you are not doing a Gibsonian ecological perception study. I think that Gibson could be used to critique our use of visualizations to accurately represent the experience of environments. Certainly this is true for critiquing static photographic visualizations.
078 Citation [27] only includes static photography and VR. It does not generate in-field data as suggested here. This has been a big problem in the literature and is one of the strengths of this paper.
094 Add (GA) after genetic algorithms.
111 In various places the term cold region, area, settlements are used. Maybe it should be cold climate? In any case, choose one term and use it consistently.
113 Should “ject Tobii” be changed to “ject. Tobii”.
113 Change “SD” to “semantic differential (SD)”. I believe this is the first instance.
119 W is the landscape visual perception score. However, W is also defined as leisure perception score (e.g., line 554). Choose what to call it. Also make it clear how it is measured—I cannot find a clear explanation.
122 The abbreviations GA and ANN have not been introduced in the text.
142 Figure 1 does not seem to be referenced in the text.
142 The text in Figure 1’s boxes must use the same terms as the article’s text and headings. I would suggest using this figure as a type of diagrammatic table of contents—add the section numbers for boxes where appropriate.
143 The figure caption should explain why some boxes are colored and other are not.
154 What term is used to consistently name these spaces? Use that here.
155 It is unfortunate that the non-snow and snow scenes are not compositionally matched—the viewpoints are different. This will impact the measurement of the factors, which impacts the results. How do you separate the seasonal from the compositional effects in the analysis? If so, this probably should be accepted as a limitation in the discussion. On the other hand, I am coming to suspect that these photos are selected from eye-tracking videos and only used for illustration purposes.
157 I am not clear how many SD pairs are used. Is Low-High used only once, or is it specific to the relevant factors. How did you pick what scales to use?
175 The questionnaire in the Supplementary Information appears to be for the space the respondent currently lives in, not the images in Table 1. Do respondents evaluate only their own residential outdoor space, or all four spaces?
177 Some of the adjective pairs in Table 2 do not agree with those listed in the questionnaire found in the Supplementary Information.
177 The bi-polar scale at the bottom of Table 2 does not match the text (e.g., line 174) or the questionnaire in the Supplementary Information.
178 I am unclear whether the participants rated photos from four sites, or rated just their own area without reference to any photos using the SD questionnaire in the Supplemental Information. I am coming to suspect that photos are not involved.
180 How are the responses distributed among the four sites?
186 It is not necessary to provide the IRB information in the text, it is specified at the end, before the references (line 789).
197 I do not understand the use of “screen” here.
202 I do not see the relevance of Gibson’s theory of Ecological Psychology here. As I understand Gibson’s theory was non-representational, which would draw into question the validity of using VR or any simulation of real environmental experiences.
205 I am confused where the “scene video” comes into this. Is it recorded by the Tobii glasses during the walk in the field, or is it observed on a monitor screen and the Tobii glasses are used to determine eye-tracking on the screen? Gibson would say these are very different experiences. For one thing, volition is not under the observer’s control when a video is used.
205 It is unnecessary to give the abbreviation (AOI) because it is only used in the text here and line 209. If it is going to be used in the figures, then it needs to be defined there too (see Instructions for Authors). Is this the same as the “factors”? What type of factors—environmental, landscape, design, …? This is another example where a consistent term needs to be used. As Leisman [17] discuss, the choice of how we move influences cognition and we lose that choice in a video.
208 I understand that “fixation, visit, glance, and saccade“ different levels of attention recorded by eye-tracking. How are they defined and interpreted? Are these the “four categories” or does that refer to something else??
212 It is not necessary to identify the IRB approval here. It is stated at the end of the paper.
233-239 It is not really necessary to describe the equipment calibration in an article for Land. This is part of the minimum criteria [40, 41] and the instructions for using the Tobii Pro Glasses 2, which are available on the internet. However, it is necessary to describe any instructions they are given or not given about how they are to walk through the site.
246 Fifty participants used a Tobii glasses during a site walk. Did they all participate in the snow and non-snow season? If so, this would be 100 possible observations, but “64 groups of effective eye-movement data” constituted a sample rate of 80% rather than 64%
255 In Figure 2f, Before Experiment, is the “Test” the “Calibration” of the equipment? If so that is what it should be called. During the Experiment, the walk is through the “Site”, not the “Scene”.
256 The caption needs to explain each part of the figure.
257 I am confused what stimulus is being used for this eye-tracking experiment. I am unclear if it is photos or videos of the site, or a computer generated 3D environment (as suggested by the use of Unity3D software).
257 What data are used to construct the 3D model in Unity?
259 Delete “(HTC, Shanghai, China)”. If it is important, cite the user manual. This could be done for Unity3D and other software also.
263 Does the use of the HTC VIVE Pro Eye “allow users to recreate realistic walking behavior”? See next comment.
269 I do not understand this sentence about validation. Is it that being able to direct movement through the space using two handle controllers is the same experience as walking? Is this supported by Leisman et al. (2016) [17]? I would think walking is an entirely different motor experience.
271 How does the experiment determine “the way that best matches the actual eye movement of the subjects”? Or is this simply an assertion without evidence?
272 What is the “motion state” of the subject?
276 Should figure 3b be Tobii Pro 2 glasses or the HTC VIVE Pro Eye headset?
276 Is the procedure truly the exact same as represented in Figures 2 and 3?
277 The caption needs to include a fuller explanation of each part of the figure. I understood there were four sites—is the top row of heatmap images from in-field walk and the bottom row from VR? Figure 3b and c suggests that only VR data are collected, but they also walk on-site with Tobii glasses—correct?
278 Figure 3c suggests that this part of the experiment is conducted indoors. Why does the time of day and sunny conditions matter? Or is it that this is what is simulated in the 3D VR model?
283 I am not sure I understand the 128 observations. There are 8 people in each of 8 scenes (4 sites in 2 seasons?). The observations are done in the field (using Tobii glasses again?) and in VR using a HIVE headset with hand controllers. Is this correct?
285 The IRB approval does not need to be mentioned here.
297 I suggest a separate paragraph about how to measure each factor and why the specific classes or thresholds are proposed. Make sure to introduce the abbreviations for each factor. This is an important part of the data used in this study and is not at all clear.
299 A citation and reference for Luranraison is needed.
299 How is the spatial aspect ratio (D/H) calculated? Is it based on the location the participant is standing? What if there are different building heights on either side of the space? What if the building height is low by the participant, but very heigh further into the view?
302 Is roof height measured for the roof nearest the viewer or by some other rule?
307 Reference [49] has nothing to do with the concept of “green viewing rate.” It is about decision trees. In any case, it is not at all certain that the area of lawn is the best measure of “green viewing,” which would normally include all types of green vegetation.
321 The implication here is that there are 49 scenarios for each season. Is this accurate, or is it 49 total?
321 I assume that each scenario is a 3D VR environment to visit by each of the 40 participants. Figure 5 indicates that it take at least 5 minutes for each scenario plus any equipment calibration time, or at least 4 hours for each participant. Is that correct?
329 I do not see how we get to 1960 “sets”. As I understand it, there are 2 sites (line 294) x 49 scenarios x 2 seasons (line 321) x 40 participants (line 326), or 3,920 observations.
330 Why are some observations not “valid”?
333 How many 3D VR environments are created? Is each one visited as described in lines 332-336?
336 It is unclear what the “visual perception rating table” is.
351 Classification normally refers to predicting categories. I understood the dependent variable to be a mean visual perception score. What are the categories?
361 It is not clear how the “visual perception score of leisure” is obtained.
366 Each equation needs to be referenced in the text.
369 How does “𝑖 represents the proportion of 𝑖”—this is saying that something represents itself. I assume that 𝑖 is the 𝑖 th instance of c classes.
370 What is “E(t)”?
377 Which equation?
382 There should be a citation for the KNN method. KNN is a classification method—what are the classes, training dataset, and prediction dataset?
394 Citation [51] does concern pedestrians, but neither “ANN,” or “neural” dsappear in the article. Citation [52] does use deep convoluted neural network, but it is unclear to this reviewer is it is the same as ANN as used in this manuscript.
400 Citation [54] does not appear to use ANN or discuss “hidden and output layers.” Is it relevant, and if so, how?
406 This seems to be the only place where separating the data into training and validations is discussed. How is this done? What is the difference between testing and validation?
416 Citations [57, 58] both concern energy use in buildings. However, [57] does not appear to use a genetic algorithm. Neither citation seems to offer guidance that would lead to the parameters presented, though I do not have the full papers.
426 What does “interest point filtering” mean? Is it identification of important factors influencing visual perception?
432 What is the meaning of “> 0.20”? Is it a reference to the probability? If so, > 0.20 seems a rather low threshold for a research paper.
467 Here and other places the order of features might be better to understand in a table than as text.
469 Change “fiction” to “fixation”.
484 Change “fiction” to “fixation”.
500 The various abbreviations need to be defined in the figures the first time they are used (see Instructions for Authors). I suggest doing it in the caption.
500 The text in the figures is too small and not legible. The Instructions for Authors suggests that all fonts be at least 8 pt.
526 Replace “cas” with “cars”?
531 I believe that this is the similarity of gaze behavior in the physical and VR landscape settings is the most important finding! However, it is not “proof” of anything. Rather it is support for accepting the validity of these 3D VR scenarios as representative of an experience in the real environment.
535 How does this study demonstrate that “actually walking in a VR environment has a more realistic effect than using a joystick remote control”? In any case, the participants did not “acutally walk” in the VR environment.
558 Five significant digits is unnecessary and misleadingly precise. At line 432 only two significant digits are used.
584 The correlation may be higher, but it is so small as to be meaningless.
619 Replaces “leaner” with “learner”.
619 Perhaps give a citation for MATLAB’s Classification Learner Toolbox?
625 How is accuracy determined?
635 Looking at Figure 16, it appears that you are using the 5-levels of the visual perception score. Is this for each individual? This seems very coarse. I not in Table 5 the mean is degraded to one of these five levels.
650 Why is the scale of the x-axis different on these two plots?
666 How do you know that “Walking on a head-mounted VR device is more realistic and closer to real-world gaze characteristics than using a joystick remote control”? This is not tested by the collected data is it?
686 A citation is needed for Qi et al.
703 It is an important finding that environmental conditions, such as temperature, ambient sound, animal life, … affect behavior and visual perceptions. These are not normally included in the VR environment.
Comments on the Quality of English LanguagePlease use terms consistently. This is a complex study and it is difficult to follow if different terms are used in different places. See the review comments for more suggestions.
Author Response
Thank you for recognizing our team's research efforts. Our team has revised and improved the manuscript according to your suggestions, please see the attached word file for the response.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsGeneral Comments
All references to line numbers is made to the revised text with No Mark-ups.
1. There need to be two additions to the Methods section. First, there is no clear mention of the dependent variable--landscape visual perception score (W)—for the first phase or VR validation (or I missed it). This variable needs to be more clearly named, as all the SD ratings are of landscape perceptions. Maybe landscape satisfaction (i.e., line 380) or preference? What is it in Chinese? Second, there is a mention of “actual interviews” in line 341, and “Talking to the subjects” in line 757. Where these interviews or informal discussions occurred during the research needs to be identified if they are to be discussed.
2. In the validation experiment, there are two methods used to navigate the VR environment: handheld paddles and a joystick controller. When the results are reported it is not clear if only the handheld paddle method of navigation is reported, or if they are averaged. If you want to say that the navigating with paddles is superior to a joystick, then separate results need to be reported.
3. Each participant in the orthogonal experiment participated for 4 hours. Was this all on the same day or spread over several sessions. That seems a long time to do VR work if one is not accustomed to it.
4. The orthogonal research design defined 49 VR scenarios for both the snow and the non-snow seasons (line 361). Figures 4 and 5 only show 16 of the total 49 scenarios, correct?
5. It is interesting to me that the analysis treats the rating scale as a five-part classifications. How does the analysis go if the 5-point scale is treated as an interval data, as it is in figures 7 through 13? You have enough cases that you could even use mean values rather than the raw ratings in the regression and discriminant analyses.
Specific Comments (by line number)
012 Change “researches” to “research”. It is not your fault—English has lots of irregular constructions.
017 Change “building elevation saturation (BS)” to “roof height difference (RHD), saturation (BS)”. Is that correct?
019 Change “apart from” to “in addition to”.
020 Change “color contrast” to “hue contrast”.
045 Change “Wen et al showed that” to “The”. The citations at the end of the sentence are for more than Wen et al.
053 Suggest changing “perceptual evaluation” to “aesthetic preference” as in line 61.
122 Change “As the result of above, this” to “This”.
124 Change “(W, it was measured by 1-5 evaluation scores 124 in the SD questionnaire below)” to “(W)”.
124 Place “it was measured by 1-5 evaluation scores 124 in the SD questionnaire below” in the Methods section where the SD scales are discussed. Make sure it is mentioned in each experiment it is measured.
129 Delete “in this study”.
148 Thank you for creating more descriptive captions.
152 Delete “environment of”.
164 Reduce the space between columns B-C and C-D so it is like columns A-B. This creates more space for the first column.
167 The SD variables are referred to as “attributes” here and “factors” in line 180. Try to be consistent. This is a complicated study and everything that can be done to make it easier to keep track of what is being done will be appreciated.
175 Change “grounds” to “ground”. Throughout the paper it is unclear what “ground” means. Is this bare earth (e.g., soil, dirt) and pavement?
179 Suggest changing “, and 15 groups of adjectives corresponding to the” to “to evaluate”.
180 I do not understand why “neutral” is needed here. Similarly, is it needed in the heading for the second column of Table 2?
181 Suggest deleting “these adjectives were partially re-peated and were all selected based on specific relevant factors”.
219 Add “(AOI)” after “areas of interest”. Or perhaps this should be done at line 79?
219 Suggest changing “ground” to “ground (e.g., bare earth and soil)”, if that is correct.
221 Delete “included fixation, visit, glance, and saccade four categories were used”.
222 Delete “. These data”.
230 Delete “on their head. This setup used the Tobii Pro Glasses 2”.
258 The use of “groups” in this paper is confusing and therefore problematic. In SPSS, I believe they use “case” to describe each row of a dataset. Maybe that is what should be used here. There are 40 subjects and two seasons when the residential landscapes are evaluated, resulting in 80 cases that are analyzed. Use the Search and Replace to make this correction, but be careful because there are some uses of “group” that do not have this meaning.
259 If the mean age for all subjects in 32.4, then it is not possible that the mean age for males is 23 and females is 27, unless there are some very old subjects who are neither male nor female.
276 What is “SU”?
277 Change “After” to “When”.
304 During the validation experiment, subjects used the HTC VIVE headset two different ways. First “actually walking,” by which I assume you mean that they use handheld paddles to simulate walking and not “actually” walk. That should be clarified. Second, they use a “remote controller,” which I understand to be a joystick. This also needs to be clarified. Use consistent terms to refer to these two forms of VR navigation.
308 There are 128 cases or responses to the experiment. 4 sites x 2 seasons x 8 local residents x 2 conditions (i.e., real site and VR?). Is this correct? If so, change “2 equal” to “2 representations” (i.e., the real site or control and the treatment or VR).
315 Suggest changing “artificial landscape” to “landscaping”. Generally, residential gardens and similar plantings are referred to as “landscaping.” “Artificial landscape” suggest something that is not biological, such a the plastic artificial grass (e.g., Astro turf) used on some sports fields.
320 Was a “correlation” calculated? If not some other term needs to be used.
322 Change “saturation” to “color saturation”. How is this determined? I do not see it explained in the Methods and is not in Table 2.
322 I suggest writing out each variable name as well as give the abbreviation in this paragraph.
327 Delete “, which has nothing to do with the position of the participant's standing position, and”.
329 Change “height” to “mean height”.
329 Delete “and if the heights of the two sides of the building are different, then take the average value of the building, and”.
341 Change “proven” to “shown”.
342 Change “wanted to investigate” to “investigated”.
343 Change “has an impact on” to “impacts”.
346 Change “PS” to “Photoshop”. Is that correct?
348 Change “to the proportion of trees, GP were” to “GP is”.
354 Change “This study” to “In this study it”.
373 How is “extreme evaluations” defined? The scale is only 1 to 5, so there is not much room for an “Extreme” response. Do you mean that all their responses were the same—e.g., all 5’s?
383 Tables 3 and 4 simply repeat the parameters discussed in section 2.4.2 Orthogonal Experiment Procedure and are probably unnecessary. These tables give the parameters but not the Orthogonal Experimental Design, which only uses a limited number of all the possible combinations. It is probably not necessary to give the parameters for each of the 49 scenarios, though they could be an appendix.
401 If 1960 cases are used, how did you determine which of the 85 snow and 97 non-snow cases to drop (see line 370)?
404 Change “environments” to “seasons”. Correct?
410 Something goes between “algorithm backpropagation”. Is it a comma?
414 ANN has not been defined in the text. You cannot depend on the abstract or figure captions for this.
465 The symbol for the Pearson correlation coefficient is “r”, not “R”.
466 Change “a correlation” to “a meaningful correlation”. Do you have a citation for the 0.20 threshold?
470 Delete “> 0.20” here and in the next couple lines.
491 Change “and cold” to “and whether they were cold”.
506 Change “in the” to “in both the”.
569 Are they walking in a “real environment” or a “VR environment” when wearing the headset?
578 Change “joystick? To “joystick to navigate”.
586 Change “actual walking VR” to “VR navigating with handheld paddles”. Is that correct?
612 Change “primary” to “linear”. Is that correct? It also occurs in lines 617,623, 628, 640, 643, and 649.
635 Add a description of the vertical dashed lines. Maybe: “The range when W is at its maximum value is shown between the vertical dotted lines.
669 Change “non-snow-season” to “non-snow season”.
682 I am not sure what data are used for this. It is not the 15% used for testing (line 438)?
707 Change “compensate for” to “ address”.
710 Change “However” to “In particular”.
784 Change “lawn” to “grass”. Also in line 789.
911 Correct the title so it is not all capital letters.
918 Correct the title so it is not all capital letters.
942 Change “kNN” to “KNN”.
966 This article was retracted. An updated version is available from the journal. It should be checked to verify that the relevant results have not changed and the reference updated.
973 Correct the title so it is not all capital letters.
Comments on the Quality of English LanguageI have suggested some edits, but they are just my suggestions for clarity.
Author Response
Thank you for your detailed revision suggestions, our team has improved the manuscript point by point based on your comments. Please see the attached file.
Author Response File: Author Response.docx
