Assessment of Urban Heat Risk in Mountain Environments: A Case Study of Chongqing Metropolitan Area, China

Round 1

Reviewer 1 Report

Thank you for submitting your manuscript to the Sustainability journal. Generally, the manuscript fits into the scope of the journal, and the structure respects Scientific Best Practice. However, there are some more comments that require revision. First of all, the manuscript does not represent an article but a case report. There are already a quite large number of case studies on urban thermal assessment.

In the introduction, you need to connect the state of the art to your manuscript goals. Tne literature review is poor. Nearly all European and American contributions on urban heat islands are missing. For that reason, the introduction requires a substantial improvement. In this way you could also solve the issue oft he quite low number of references in the current version of the manuscript. Please follow the literature review by a clear and concise state of the art analysis. This should clearly show the knowledge gaps identified and link them to your manuscript goals. Please reason both the novelty and the relevance of your manuscript goals. In the end of the introduction section you should clearly explain the gaps from the literature and conclude the need to close the gaps. In this frame you need to explain the scope of the work presented.

In the methodology section must be indicated the source of all figures. The methodology section is well written. However there are some questions. What is the uncertainty of the modeling procedures ? PLease explain how the model was calibrated and validated.

In the conclusions, in addition to summarising the actions taken and results, please strengthen the explanation of their significance. It is recommended to use quantitative reasoning comparing with appropriate benchmarks, especially those stemming from previous work.

I recommend for the conclusions to make a reference to the subject of green Infrastructure as nature-based solution to buffer urban heat islands.

Author Response

Part 1: Response to Reviewer #1

Issue 1: Thank you for submitting your manuscript to the Sustainability journal. Generally, the manuscript fits into the scope of the journal, and the structure respects Scientific Best Practice. However, there are some more comments that require revision. First of all, the manuscript does not represent an article but a case report. There are already a quite large number of case studies on urban thermal assessment.

Response: Accepted and revised.

First, we appreciate the positive comments on the manuscript. In the process of revision, we considerably revised the introduction, methodology, and application significance of the manuscript, trying to enhance the scientific significance of the manuscript and to avoid it looking like a case report.

Issue 2: In the introduction, you need to connect the state of the art to your manuscript goals. The literature review is poor. Nearly all European and American contributions on urban heat islands are missing. For that reason, the introduction requires a substantial improvement. In this way you could also solve the issue of the quite low number of references in the current version of the manuscript. Please follow the literature review by a clear and concise state of the art analysis. This should clearly show the knowledge gaps identified and link them to your manuscript goals. Please reason both the novelty and the relevance of your manuscript goals. In the end of the introduction section you should clearly explain the gaps from the literature and conclude the need to close the gaps. In this frame you need to explain the scope of the work presented. 

Response: Accepted and revised.

According to the reviewer's opinion, we revised the introduction. In fact, in the original manuscript, we do analyze in detail the current research methods (the second paragraph), shortcomings and objectives of this study (the third paragraph). But according to this suggestion, we find that the above parts are scattered and unclear. Therefore, we add the relevant research on urban thermal environment risk through literature review and show the knowledge gaps identified and link them to our research objectives. At last, we rearranged the research objectives to make them clearer. The main changes are as follows.

At present, the researches on urban thermal environment risk focus on the mapping of thermal environment risk, especially in the context of global climate change and heat wave. Researchers usually grade and evaluate the air temperature monitored by meteorological stations and the LST observed by remote sensing from the perspective of climate vulnerability or human exposure. In this process, researchers did use some vulnerability and risk index indexes, such as manual indicator removal, and more complicated techniques, such as Monte Carlo simulation and variance-based global sensitivity analysis. However, the climate background and development of each city are different, the above thermal environment analysis framework based solely on temperature itself is not applicable to all cities. In addition, these thermal environment risk mapping and analysis mostly reproduce the historical situation and do not consider the interaction of surrounding pixels. Now, there is still a lack of universal evaluation criteria of urban thermal environmental risk. It is worth noting that land use carries almost all human activities and energy balance process. Analysis based on Cellular automata-Markov model offer an opportunity for simulate and forecast land use change, providing insight into future LST characteristics and heat risk based on land use. Establishing the spatial relationship between LUCC and urban thermal environment risk can quickly predict the possible future risks.

As noted above, our objective is (1) to construct a spatial prediction model of urban thermal environmental based on spatial relationship between LUCC and LST; (2) to put forward an evaluation criteria of urban thermal environmental risk based on LST grades in the present period and the predict period. This method could provide guidance to the smart urban growth, and decision support for urban planning and urban ecological security so as to prevent and control urban thermal environment risk.

Issue 3: In the methodology section must be indicated the source of all figures. The methodology section is well written. However, there are some questions. What is the uncertainty of the modeling procedures? Please explain how the model was calibrated and validated. 

Response: Accepted and revised.

In the revised manuscript, we added the accuracy verification and uncertainty of the model. In the original manuscript, we calculated the consistency between the predicted LST data and the observed LST data by kappa coefficient, and in the revised version, we add the formula of kappa coefficient to make the calculation process clearer. In additional, we further analyze the possible uncertainties of the model. The detailed modification is as follows.

Through the coupled Markov-CA model, the LST can be predicted. It is necessary to further compare the difference between the predicted LST results and the real observed LST to determine the accuracy of the coupled Markov-CA model. Because kappa coefficient can test the consistency of prediction results and observation data, this study compares the two data through kappa coefficient.

where r is the number of rows in the confusion matrix between the prediction LST results and the observation LST results; x_ii is the number of along the diagonal; x_i+ is the total number of row i; x_+i is the total number of column i; and N is the total number of cells.

The value range of kappa coefficient is [-1,1], usually between [0,1]. The larger the kappa coefficient is, the better the consistency between the prediction results and the observation results is. In this study, the biggest uncertainty lies in the mean standard deviation method of LST. Because of the interannual difference, there is a certain classification error for LST. These errors will also be transferred to the Markov-CA model prediction process. In addition, with the increase of urban construction intensity, the LST difference between source and sink of urban thermal environment is becoming larger and larger, which also interferes with the simulation results.

Issue 4: In the conclusions, in addition to summarizing the actions taken and results, please strengthen the explanation of their significance. It is recommended to use quantitative reasoning comparing with appropriate benchmarks, especially those stemming from previous work. I recommend for the conclusions to make a reference to the subject of green Infrastructure as nature-based solution to buffer urban heat islands.

Response: Accepted and revised.

In the revised manuscript, we improved the conclusion based on the suggestions of the reviewers, mainly focusing on the research significance and promotion. In combination with the most popular national strategy, national territory development plan, in China, this paragraph expounds the necessity of the research and puts forward possible suggestions. The green infrastructure mentioned by the reviewer is applied to the proposal in the form of minimum ecological safety distance, which avoids the appearance of large plaques with high-risk and extremely high-risk for urban thermal environment. The detailed modification is as follows.

In this paper, the UTERM model has been built to predict the urban thermal environmental risk. The MARKOV-CA rule set of variety of urban thermal environment is established based on the spatial relationship between LST grades and land use type. And then based on the new developed classification criteria of thermal environmental risk, the urban thermal environmental risk could be accurately predicted by the LST grades and LUCC in the process period and the predict period, which could provide guidance to the smart urban growth in order to cope with extreme climatic risk caused by disordered exploitation. The results showed that the risk of urban thermal environment might be increasing in Chongqing metropolitan area. It showed that the possibility of risk has transited to higher levels, particularly on the thermal environment risk ecotone, and the area of extreme risk zone had increased by more than 10%. Based on the research results, the paper further proposes the urban thermal environment spatial control measures. Spatially, urban construction land showed remarkable performances in the rules of transform into the higher LST grade and the highest LST grade. Urban construction land has a very high spatial correlation with extreme risk area and high-risk area, so it is important to optimize the spatial allocation of various land-use types and set up ecological corridors dividing the larger urban patch to prevent urban sprawl in order to reduce the risk of urban thermal environment. Considering urban management, we should strictly control the disordered growth of urban construction land, and in combination with the requirements of national territory development plan, optimize the urban growth boundary from the perspective of thermal environment, take different development control measures for the areas (including development zones to be optimized, key development zones, limited development zones, and prohibited development zones) with different thermal environment risk levels. In the safety urban thermal environment area, urban growth potential can be explored moderately. However, in the high-risk area, there should be strict control measures to prohibit high-intensity urban development and further put forward ecological cooling measures. Especially for the current urbanization tends to double (multi) nuclear city and urban agglomeration trends, it is significant for reducing the possibility of urban thermal environment risk, under the premise of avoiding the appearance of large plaques with high-risk and extremely high-risk for urban thermal environment, to measure the minimum ecological safety distance between cities, establish ecological corridors and ecological networks and delimit urban growth boundary.

Reviewer 2 Report

General comments

All issues were addressed below.

Other comments are as follows:

Line 1: Please add “China” at the end of the title because not everyone knows where it is Chongqing Metropolitan Area. E.g. : “Assessment of Mountain Urban Thermal Environment Risk: A Case Study of Chongqing Metropolitan Area, China”. Lines 55-90: This paragraph is too large and difficult to read; please divide it into 2-3 relevant paragraphs. Line 110: Insert within Figure 1. a general map with the geographical location of the Chongqing Metropolitan Area in China. Line 235: Figure 5. Please use a color palette specific to climate maps as you used in Figure 7. Line 328. Because you have used many abbreviations, you can add a list of abbreviations at the end of the article. Here https://www.mdpi.com/2073-4441/11/9/1832/htm you can find an example (see Abbreviations section after Conflicts of Interest section). Lines 330-411. I think you can add more references in the context of a very well documented topic related to climate risk assessment.

Author Response

Part 2: Response to Reviewer #2 

Issue 1: Line 1: Please add “China” at the end of the title because not everyone knows where it is Chongqing Metropolitan Area. E.g. : “Assessment of Mountain Urban Thermal Environment Risk: A Case Study of Chongqing Metropolitan Area, China”. 

Response: Accepted and revised.

We revised the title as “Assessment of Mountain Urban Thermal Environment Risk: A Case Study of Chongqing Metropolitan Area, China”.

Issue 2: Lines 55-90: This paragraph is too large and difficult to read; please divide it into 2-3 relevant paragraphs.

Response: Accepted and revised.

According to the reviewer's suggestion, we reorganized the second paragraph into two new paragraphs in the revised manuscript. The first paragraph compares the applicable conditions and advantages of various methods in urban thermal environment research. In the second paragraph, the shortcomings of various methods are compared, which further leads to the purpose of this study. The revised manuscript is as follows.

Researchers have endeavored to employ theoretical and technical methods-such as statistical model, energy balance model, numerical model, analytical model and physical model-to probe into the relationship between some characterization index such as land use and cover change (LUCC) and spatio-temporal patterns and processes of urban thermal environment. Statistical models often use mathematical statistics to study the relationship between UHI and urban scale, population density, socioeconomic factors, and establish corresponding linear equations, so as to predict urban thermal environment processes. In addition, a variety of mathematical methods, such as grey system theory, Markov chain model, neural network are applied to urban thermal environment prediction and simulation. Energy balance model targets mainly at revealing the mechanism of urban heat distribution and formation; It marks the focus of studies in the thermal environment shifting from description of the phenomenon to explanation of the mechanism, and the model is used to study the influencing factors by transfer and exchange of thermal between urban and surrounding areas. The numerical model which consist of atmospheric boundary layer meteorological model, urban canopy model and CFD (Computational Fluid Dynamics) simulation technology is a mechanism model analyzing surface energy balance and temporal and spatial variation of temperature field based on thermodynamics and dynamics. The numerical model can calculate and reveal the process of the generation, development, and change of the UHI in different time and space which makes up for the deficiency of the statistical model and the energy balance model. The analytic model is a process of optimizing the nonlinear differential equations of the numerical model and reestablishing the model, which could visually reveal the fluctuation properties of UHI. The most common model in the physical model is the wind tunnel test, which reduces the study area to a certain proportion and simulates the gas and heat exchange process. It provides a more intuitive approach to analyze changes of UHI and to analyze the impact of different parameters on the UHI.

Furthermore, there are still many improvements for these models in the study of urban thermal environment. These statistical models are just analysis of phenomenon but fail to reveal the mechanism of spatiotemporal change of urban thermal environment. Moreover, since the developmental process of any city could not be identical, the analysis would be somehow interfered. Therefore, the statistical model is usually limited in a specific research area but difficult to promote to a wide range. The energy balance model requires too many parameters, and its focus on formation mechanism of micro-regional urban thermal environment make it difficult to describe the process of the macro-regional spatial and temporal changes. For the numerical model, the unique underlying surface properties (such as urban man-made heat sources, urban building canopy) and their terrain characteristics in the study area will make the simulation results unusually sensitive. Similarly, the analytic model needs to assume many conditions and therefore inevitably affect the analysis results. More directly, the physical model is time-consuming and costly. So how to establish a spatio-temporal pattern prediction model of urban thermal environment based on land use types becomes the key to rapidly identify urban thermal environment risks.

Issue 3: Line 110: Insert within Figure 1. a general map with the geographical location of the Chongqing Metropolitan Area in China. 

Response: Accepted and revised.

In the revised figure 1, we added the location of the research area in China. The revised figure is as follows.

Figure 1. The elevation and LUCC of Chongqing metropolitan area in 2015

Issue 4: Line 235: Figure 5. Please use a color palette specific to climate maps as you used in Figure 7. 

Response: Accepted and revised.

According to the reviewer's suggestion, we adjusted the color of Figure 5. The revised figure is as follows.

Figure 5. The probability of each LST grade in Chongqing metropolitan area in 2015. (a) Probability of lowest temperature level; (b) Probability of lower temperature level; (c) Probability of medium temperature level; (d) Probability of higher temperature level; (e) Probability of highest temperature level

Issue 5: Line 328. Because you have used many abbreviations, you can add a list of abbreviations at the end of the article. Here https://www.mdpi.com/2073-4441/11/9/1832/htm you can find an example (see Abbreviations section after Conflicts of Interest section).

Response: Accepted and revised.

According to the reviewer's suggestion, we adjusted the color of Figure 5. The revised figure is as follows.

LUCC      Land cover change

LST         Land surface temperature

UTERM     Urban Thermal Environment Risk Model

UHI         Urban heat island

CA          Cellular automata

MODIS      Moderate resolution imagine spectroradiometer

Issue 5: Lines 330-411. I think you can add more references in the context of a very well documented topic related to climate risk assessment. 

Response: Accepted and revised.

We have added some references on climate risk assessment in the discussion and conclusion section.

Author Response File: Author Response.pdf

Reviewer 3 Report

The submitted manuscript is a study about the relation of land use and land surface temperature in Chongqing metropolitan area. A cellular automata (CA) – Markov model was trained with land use data and land surface temperatures to make predictions about the possible future development of land surface temperature which was used for a categorization into different thermal risk classes.

A number of problems is associated with the manuscript in the current form. In the introduction, the originality is not clarified. First of all, the acronym CA is introduced nowhere in the paper but has to be explained briefly. Secondly, there are at least two papers published (Firozjaei et al., 2018, Mushore et al., 2017) that use the same approach to study the potential future development of land surface temperature in relation with land cover development. You just mention one of the papers and you do not explain in which way your work goes beyond what has been done before.

Another major problem is that you make a risk assessment such as: “thermal disasters occur frequently” for the highest temperature level but you do not provide any scientific basis that underlines your statement. First, you do not define what you mean with “thermal disasters” but the term disasters is normally related with human deaths and “frequently” would normally mean at least several times in a year in my understanding. Common sense tells us that this is not the case, so there is a major conceptional and scientific gap in the paper. You also make the statement that the lowest temperature level is associated with “ideal safety” and “thermal disaster has not occurred”. Looking at the map this holds mostly for areas that are >2000 m above sea level and eventually get pretty cold in winter which could lead to cold deaths. What I want to say by that is that the terminology is wrong. This paper is not about “thermal risk” but about “heat risk”, which is an important difference.

In summary and for the above mentioned reasons I do not think the manuscript is suitable for publication in an international peer-reviewed journal such as Sustainability.

Author Response

Part 3: Response to Reviewer #3 

Issue 1: The submitted manuscript is a study about the relation of land use and land surface temperature in Chongqing metropolitan area. A cellular automata (CA) – Markov model was trained with land use data and land surface temperatures to make predictions about the possible future development of land surface temperature which was used for a categorization into different thermal risk classes. A number of problems is associated with the manuscript in the current form. In the introduction, the originality is not clarified. First of all, the acronym CA is introduced nowhere in the paper but has to be explained briefly.

Response: Accepted and revised.

In the revised manuscript, we almost rewrote the introduction. The advantages and disadvantages of various research methods are compared, and the research objectives and innovation of this study are clearly put forward. Of course, the revised version also gives a brief introduction to Cellular automata-Markov model. See the introduction for details.

As noted above, our objective is (1) to construct a spatial prediction model of urban thermal environmental based on spatial relationship between LUCC and LST; (2) to put forward an evaluation criteria of urban thermal environmental risk based on LST grades in the present period and the predict period. This method could provide guidance to the smart urban growth, and decision support for urban planning and urban ecological security so as to prevent and control urban thermal environment risk.

Issue 2: Secondly, there are at least two papers published (Firozjaei et al., 2018, Mushore et al., 2017) that use the same approach to study the potential future development of land surface temperature in relation with land cover development. You just mention one of the papers and you do not explain in which way your work goes beyond what has been done before. 

Response: Accepted, clarify and revised.

I have carefully read the two papers recommended by the author. Compared with these two articles, the objective of the research is to assess and predict the risk, rather than the values of the LST and urban heat island intensity. However, the climate background and development degree of each city are different, the above thermal environment risk analysis framework based solely on temperature itself is not applicable to all cities. The LST and urban heat island intensity is just only a number, but the risk is more closely related to the reality living of human beings. In this article, we further put forward an evaluation criterion of urban thermal environmental risk based on LST grades in the present period and the predict period. After the assessment of urban heat risk, urban planning and regulation could be put forward more scientifically to improve the urban climate environment. We also supplement this improvement in the revised manuscript. See the introduction for details.

At present, the researches on urban thermal environment risk focus on the mapping of thermal environment risk, especially in the context of global climate change and heat wave. Researchers usually grade and evaluate the air temperature monitored by meteorological stations and the LST observed by remote sensing from the perspective of climate vulnerability or human exposure. In this process, researchers did use some vulnerability and risk index indexes, such as manual indicator removal, and more complicated techniques, such as Monte Carlo simulation and variance-based global sensitivity analysis. However, the climate background and development of each city are different, the above thermal environment analysis framework based solely on temperature itself is not applicable to all cities. In addition, these thermal environment risk mapping and analysis mostly reproduce the historical situation and do not consider the interaction of surrounding pixels. Now, there is still a lack of universal evaluation criteria of urban thermal environmental risk. It is worth noting that land use carries almost all human activities and energy balance process. Analysis based on Cellular automata-Markov model offer an opportunity for simulate and forecast land use change, providing insight into future LST characteristics and heat risk based on land use. Establishing the spatial relationship between LUCC and urban thermal environment risk can quickly predict the possible future risks.

Issue 3: Another major problem is that you make a risk assessment such as: “thermal disasters occur frequently” for the highest temperature level but you do not provide any scientific basis that underlines your statement. First, you do not define what you mean with “thermal disasters” but the term disasters is normally related with human deaths and “frequently” would normally mean at least several times in a year in my understanding. Common sense tells us that this is not the case, so there is a major conceptional and scientific gap in the paper. You also make the statement that the lowest temperature level is associated with “ideal safety” and “thermal disaster has not occurred”. Looking at the map this holds mostly for areas that are >2000 m above sea level and eventually get pretty cold in winter which could lead to cold deaths. What I want to say by that is that the terminology is wrong. This paper is not about “thermal risk” but about “heat risk”, which is an important difference. 

Response: Accepted, clarify and revised.

First, I quite agree with the reviewer's point of view on winter climate. Chongqing is one of the three most famous furnaces in China. Due to the terrain, the temperature is particularly high in summer, which has an impact on the reality living of human beings. Therefore, this study focuses on the summer thermal environment in Chongqing. We searched the relevant research and news and did not find the report that people died of freezing in summer in Chongqing. On the contrary, these areas with an altitude of more than 2000 meters are regarded as summer tourism destinations. It is necessary to protect the ecological environment of these areas for summer vacation. Therefore, in the initial design of this paper, the thermal environment risk of these areas is defined as the ideal safety area of thermal environment.

Second, the classification criteria and significance of all thermal environment risks are shown in Table 3.

Third, according to the reviewer's opinion, we change thermal to heat in the revised manuscript.

Round 2

Reviewer 1 Report

Thank you for providing the revised version. The majority of my concerns was considered. I still have some further comments: the section Discussion and Conclusion must be separated into two sections to make clear what are the conclusions.

The conclusions are still weak and should be improved through being more detailed in terms of quantitative reasoning comparing with appropriate benchmarks, and conclusions of the improvements compared with previous work. Also the reference to green Infrastructure as nature-based solution to buffer urban heat islands ist still weak.

Author Response

Part 1: Response to Reviewer #1 Issue 1: Thank you for providing the revised version. The majority of my concerns was considered. I still have some further comments: the section Discussion and Conclusion must be separated into two sections to make clear what are the conclusions. The conclusions are still weak and should be improved through being more detailed in terms of quantitative reasoning comparing with appropriate benchmarks, and conclusions of the improvements compared with previous work. Also the reference to green Infrastructure as nature-based solution to buffer urban heat islands is still weak. Response: Accepted and revised. According to the reviewer's opinions, we divided the discussion and conclusion into two parts. In the discussion, we compare the difference between the new UHERM model and the previous model from two aspects of availability and applicability. In terms of applicability, especially, we have increased the discussion on the mutual support of the new UHERM model and green infrastructure in detail. The detailed modification is as follows. 5 Discussion In this paper, a UHERM model was constructed to predict urban heat environment risk. A CA-Markov rule set for variation in the urban heat environment was established based on the spatial relationship between LST grade and land use type. In further, using the newly developed classification criteria for heat environment risk, the urban heat environment risk could be accurately predicted from LST grades and LUCC in the processing and prediction periods, which could provide guidance for the smart urban growth in response to the extreme climatic risk caused by overexploitation. Compared with the existing urban heat environment prediction model, the new model has higher availability. First, the model does not need to consider the absolute LST, rather than reclassifies the LST into highest temperature, higher temperature, medium temperature, lower temperature, and lowest temperature using the mean-standard deviation method, which avoids the spatio-temporal influences caused by climate type, special terrain and seasonal change. Second, the prediction of urban heat risk is based on the relationship between LST and LUCC in the UHERM model. Most of the previous models predicted the risk of urban heat environment based on population density, socioeconomic status, and achieved good results [41, 42]. However, most of the data in these models, calculated by empirical formula or anthropogenic statistics, cannot represent all the factors that lead to the change of urban heat environment. In the UHERM model, land use carries all human production and living, even the global climate change effect. Most importantly, LUCC could be accurately interpreted by remote sensing satellites. These advantages can also avoid the influence of spatio-temporal differences on model accuracy. In addition, the UHERM model is more applicable because of the new developed evaluation criteria of UHI risk. Based on the present results, we further propose urban heat environment spatial control measures. Spatially, urban developed land exhibited remarkable changes, shifting into the higher and highest LST grades. Urban developed land has a very high spatial correlation with the extreme risk and high risk areas, and therefore, it is important to optimize the spatial allocation of various land use types and maintain ecological corridors dividing large urban areas to prevent urban sprawl and reduce the risk of UHI effects. Urban management should strictly control the expansion of urban developed land, and in combination with the requirements of the national territory development plan, optimize urban growth boundaries from the perspective of the heat environment and implement different development control measures for areas (including development zones that should be optimized, key development zones, limited development zones, and prohibited development zones) with different heat environment risk levels. In the safety grade area, urban growth potential can be explored moderately. However, in the high-risk area, there should be strict control measures to prohibit high-intensity urban development and further ecological cooling measures should be implemented. Considering the current urbanization trend toward double (or multi) nuclear cities and urban agglomeration areas, it is essential to minimize the UHI risk. This can be accomplished by avoiding the formation of large contiguous areas of high risk and extreme high risk in the urban environment, measuring the minimum ecological safety distance between cities, establishing ecological corridors and ecological networks, and delimiting urban growth boundaries [34]. Therefore, the spatial identification of different levels of urban heat environment risk zone and the spatial-temporal change analysis of topological relationship among the risk zones will help to improve the scientific of green infrastructures spatial planning [44, 45]. Efficient green infrastructure as nature-based solutions will be of great significance to mitigate the risk of urban heat environment and maintain the sustainable development of the city [46]. 6 Conclusions The UHI risk model described here has high simulation and prediction accuracy. The results showed that urban heat environment risk may be increasing in the Chongqing metropolitan area. The risk level shifted toward higher levels, particularly on the heat environment risk scale, and the area of covered by the extreme risk zone increased by more than 10%. For sustainable development and the prevention of UHI risks, it is essential to predict the areas at high risk of future UHI effects and identify the formation of new patches of high risk [6, 8-10]. However, this method has weaknesses that can be improved in the future. First, the spatial resolution of MODIS LST products and LUCC data is 1 km, which greatly affects prediction accuracy for small- and medium-sized cities [47, 48]. In addition, the CA-Markov model only couples the first-level classification of LUCC with LST. However, the internal structure of LUCC would affect surface energy balance processes, such as the influence of the forest canopy on surface roughness [3, 4, 49], the influence of soil moisture on surface albedo and energy exchange processes [16, 20, 50], and the influence of building form and spatial layout on the urban ventilation environment [51]. Therefore, the spatial resolution of LST products and LUCC data should be improved, and the spatial relationship between LST and LUCC should be further explored. Second, natural factors such as topography and the local background climate affect the prediction of UHI effects [52]. Therefore, for mountainous cities, it is essential to determine the impact of complex terrain on LST [53]. Third, further consideration should be given to the impacts of socio-economic factors such as anthropogenic heat on the urban heat environment [54] to improve the accuracy and precision of UHI risk prediction. References 41.Zhang, W.; Zheng, C.; Chen, F. Mapping heat-related health risks of elderly citizens in mountainous area: A case study of Chongqing, China. Sci Total Environ. 2019, 663, 852-866, doi:10.1016/j.scitotenv.2019.01.240. 42. Räsänen, A.; Heikkinen, K.; Piila, N.; Juhola, S. Zoning and weighting in urban heat island vulnerability and risk mapping in Helsinki, Finland. Reg. Environ. Change. 2019, 19(5), 1481-1493, doi:10.1007/s10113-019-01491-x. 43. Coppola, E.; Rouphael, Y.; De Pascale, S.; Moccia, F.D.; Cirillo, C. Ameliorating a complex urban ecosystem through instrumental use of softscape buffers: proposal for a green infrastructure network in the metropolitan area of Naples. Frontiers. Plant. Sci. 2019, 10, doi:10.3389/fpls.2019.00410. 44. Bartesaghi-Koc, C.; Osmond, P.; Peters, A. Mapping and classifying green infrastructure typologies for climate-related studies based on remote sensing data. Urban. Form Urban. Gree. 2019, 37, 154-167, doi:10.1016/j.ufug.2018.11.008. 45. Marando, F.; Salvatori, E.; Sebastiani, A.; Fusaro, L.; Manes, F. Regulating ecosystem services and green infrastructure: assessment of urban heat island effect mitigation in the municipality of Rome, Italy. Ecol. Model. 2019, 392, 92-102, doi:10.1016/j.ecolmodel.2018.11.011. 46. Makido, Y.; Hellman, D.; Shandas, V. Nature-based designs to mitigate urban heat: the efficacy of green infrastructure treatments in Portland, Oregon. Atmos. 2019, 10(5), 282, doi:10.3390/atmos10050282.

Reviewer 3 Report

The submitted manuscript is a study about the relation of land use and land surface temperature in Chongqing metropolitan area. A cellular automata (CA) – Markov model was trained with land use data and land surface temperatures to make predictions about the possible future development of land surface temperature which was used for a categorization in to different thermal risk classes.
There are still improvements to be made which I explain in detail in the remarks given below.

Remarks:
The English needs to be improved, I recommend to use an English language service for that. A space is needed between numbers and units. You changed your model name from UTERM to UHERM but this is not updated throughout the manuscript including the abstract itself. The acronym “CA” is not introduced in the abstract. Rename Section 3 “Method” to “Methodology”. I would also rename “Study areas and materials” to “Study area and materials”.

Introduction:

In the 2nd paragraph (ll. 56-88) you give a lengthy overview of all the approaches used in urban climatology (statistical models, CFD models, energy balance models etc.). I do not think this is necessary and I recommend to shorten this to 1-2 sentences at most. Instead of introducing all the approaches that are used in the entire field, the introduction should focus on the approach that is used in this study and highlight the literature that has used this approach before. Then make a clear statement in which way this study goes beyond the earlier studies and what the motivation of the study was (heat risk detection and characterization in Chongqing I guess).

Methodology:
In section 3.3 your risk levels and the “significance of heat environment risk level” are still not based on scientific evidence. For example, for the “Extreme risk” level you write “heat disasters usually”. You do not define what that means at all. Generally risk from heat should be related to an increase in heat illness and/or mortality, which is nowhere to be found in this manuscript. I did a quick search on “Chongqing heat mortality” and found at least 4 publications which study the relationship between heat and health for your area of interest, which you can use to justify your risk levels. But I strongly recommend that you rewrite the “significance of heat environment risk level” in Table 3. For the highest risk level, you could for example write “Potentially highest risk of heat related illness, morbidity and mortality increases” or something similar. Furthermore you will have to mention that these studies commonly study the relationship of health with air temperature which is not equal to land surface temperature which is studied here. But there are a number of studies showing that there is high agreement between increased levels of LST and air temperature, which is something you need to mention as well then.

Discussion and Conclusion
You state that you developed a new model, which you name UHERM. You need to add a statement here about the availability of that model. Also, you need to discuss whether or not this model can be readily applied in other regions as well.

Detailed remarks:

l. 27: introduction for acronym CA is missing

l. 116: you mention the cellular automata model here for the first time, then you also need to add “(CA)” to introduce this acronym

ll. 131-132: wording needs improvement

eq. (3): should it be written C(t+a) = P x C(t) instead of having t+a in the exponent?

l. 221: the term heat disaster is undefined

Section title 4.1: change “Based” to “based”

Table 4: which year is represented by columns and which one by rows?

l. 335: check the wording of “disordered exploitation”

l. 336: change “transited” to “transitioned”

Author Response

Part 2: Response to Reviewer #3 

Remarks 1: The English needs to be improved, I recommend to use an English language service for that. A space is needed between numbers and units. You changed your model name from UTERM to UHERM but this is not updated throughout the manuscript including the abstract itself. The acronym “CA” is not introduced in the abstract. Rename Section 3 “Method” to “Methodology”. I would also rename “Study areas and materials” to “Study area and materials”. 

Response: Accepted and revised.

The English in this document has been checked by at least two professional editors, both native speakers of English. For a certificate, please see:

http://www.textcheck.com/certificate/AAVWxb

Issue 1: Introduction: In the 2nd paragraph (ll. 56-88) you give a lengthy overview of all the approaches used in urban climatology (statistical models, CFD models, energy balance models etc.). I do not think this is necessary and I recommend to shorten this to 1-2 sentences at most. Instead of introducing all the approaches that are used in the entire field, the introduction should focus on the approach that is used in this study and highlight the literature that has used this approach before. Then make a clear statement in which way this study goes beyond the earlier studies and what the motivation of the study was (heat risk detection and characterization in Chongqing I guess). 

Response: Accepted and revised.

According to the reviewer's suggest, we rewrite the introduction. We have deleted many introductions about urban thermal environment methods. In the new revised introduction, we mainly focus on the shortcomings of the research on thermal environment risk, which further led to the objective of our research.

The detailed modification is as follows.

Introduction 

The urban heat island (UHI) effect is a product of urbanization; this term refers to the phenomenon of temperature in an urban area being significantly higher than in the suburbs [1]. The rapid increase in land surface temperature (LST) and increased UHI are result of primarily from urban developed land expanding to encroach upon cultivated land, grassland, and forest land [2, 3]. During the process of urbanization, constant changes in underlying surface features and land use cover result in decreased latent heat flux and the increased sensible heat flux, which are the root causes of an increased UHI effect [4]. At present, the urban heat problem is no longer a simple climate and environmental issue, but a major risk to the sustainable development of the urban environment [5]. Therefore, it is necessary to reveal the mechanism through which land use type affects the spatio-temporal patterns of urban heat environments to predict the risk of urban heat effects and effectively control the deterioration of the urban heat environment.

At present, research on urban heat environment risk focuses on mapping heat risk and vulnerability in various environments [5-7], especially in the context of global climate change and heat waves [8-10]. Researchers have endeavored to employ theoretical and technical methods, such as statistical [11-19], energy-balance [3,4,20], numerical [21-23], analytical [24], and physical models [25]. For this process, researchers usually grade and evaluate air temperature data from meteorological stations and the LST data observed via remote sensing from the perspective of climate vulnerability or human exposure [26], and developed some vulnerability and risk indexes, such as manual indicator removal [27], as well as more complicated techniques, such as Monte Carlo simulation and variance-based global sensitivity analysis [28]. However, the climate background resulting from special environment or terrain and development pattern of each city are different, and the heat environment analysis framework described above, which is based solely on temperature, is not applicable to all cities. In addition, mapping and analysis studies of heat environment risk generally reproduces the historical situation and do not consider interactions with the surrounding pixels. Universal evaluation criteria for urban heat environmental risk remain lacking today. Notably, land use and cover change (LUCC) is associated with almost all human activities and energy balance processes. Establishing a spatio-temporal pattern prediction model or the urban heat environment based on land use type is essential to rapidly identifying urban heat environment risks.

An analysis based on the cellular automata-Markov (CA-Markov) model offers an opportunity to simulate and forecast land use changes, providing insight into future LST characteristics and heat risks based on land use [18, 29]. Establishing the spatial relationship between LUCC and urban heat environment risk will allow for the prediction of possible future risks. As noted above, our objectives are (1) to construct a spatial prediction model of the urban heat environmental based on the spatial relationship between LUCC and LST; and (2) to establish evaluation criteria for urban heat environmental risk based on LST grades for the present and a prediction period in mountain city with high temperature risk. This method could provide guidance for smart urban growth, urban planning, and urban ecological security to minimize and control environmental risk from urban heat.

Issue 2: Methodology: In section 3.3 your risk levels and the “significance of heat environment risk level” are still not based on scientific evidence. For example, for the “Extreme risk” level you write “heat disasters usually”. You do not define what that means at all. Generally risk from heat should be related to an increase in heat illness and/or mortality, which is nowhere to be found in this manuscript. I did a quick search on “Chongqing heat mortality” and found at least 4 publications which study the relationship between heat and health for your area of interest, which you can use to justify your risk levels. But I strongly recommend that you rewrite the “significance of heat environment risk level” in Table 3. For the highest risk level, you could for example write “Potentially highest risk of heat related illness, morbidity and mortality increases” or something similar. Furthermore you will have to mention that these studies commonly study the relationship of health with air temperature which is not equal to land surface temperature which is studied here. But there are a number of studies showing that there is high agreement between increased levels of LST and air temperature, which is something you need to mention as well then.

Response: Accepted and revised.

In the revised manuscript, based on the selected references, we justify the definition of urban heat environment risk. The revised definition of urban heat environment risk seems more scientific. See Table 3 for detailed modification.

Of course, as the reviewer said, previous studies focused on the relationship between the air temperature and heat related illness, but there was a high consistency between air temperature and LST. We have also added these contents in this part.

The detailed modification is as follows.

It has been shown previously that there is a strong relationship between air temperature and the health of both humans and the ecosystem [34-36]. Heat waves, which are caused by continuous extreme weather, are one of the most important factors in this relationship [37, 38]. Long-term exposure to high temperature will greatly increase mortality. Although, these above studies commonly defined the relationship of health with air temperature, there are a number of studies showing that there is high agreement between increased levels of LST and air temperature [37, 38]. Therefore, this paper uses LST levels in summer to define the urban heat environment risk. The classification criteria for heat environment risk were established based on LST grade during the processing and prediction periods (Figure 3 and Table 3).

Table 3. Evaluation criteria for urban heat environmental risk.

The combination rule was adopted for heat environment grades for the processing and prediction years. If a pixel met two risk level definitions at the same time, the higher risk level was selected.

References

Basu, R. High ambient temperature and mortality: a review of epidemiologic studies from 2001 to 2008. Environ. Health-Glob. 2009, 8(1), 40, doi:10.1186/1476-069X-8-40. Patz, J. A.; Campbell-Lendrum, D.; Holloway, T.; Foley, J. A. Impact of regional climate change on human health. Nature. 2005, 438(7066), 310-317, doi:10.1038/nature04188. Carnicer, J.; Coll, M.; Ninyerola, M.; Pons, X.; Gerardo, S.; Mooney, J. P. A. Widespread crown condition decline, food web disruption, and amplified tree mortality with increased climate change-type drought. P. Natl. Acad. Sci. USA. 2011, 108(4), 1474-1478, doi:10.2307/41001885. Macintyre, H.; Heaviside, C.; Taylor, J.; Picetti, R.; Vardoulakis, S. Assessing urban population vulnerability and environmental risks across an urban area during heatwaves – Implications for health protection. Sci. Total. Environ. 2017, 610-611, 678-690, doi:10.1016/j.scitotenv.2017.08.062. Mcgeehin, M. A.; Mirabelli, M. The potential impacts of climate variability and change on temperature-related morbidity and mortality in the United States. Environ. Health. Persp. 2001, 109(2), 185-189, doi:10.1289/ehp.109-1240665.

Issue 3: Discussion and Conclusion. You state that you developed a new model, which you name UHERM. You need to add a statement here about the availability of that model. Also, you need to discuss whether or not this model can be readily applied in other regions as well. 

Response: Accepted and revised.

According to the reviewer's suggestions, we divided the discussion and conclusion into two parts. In the discussion, we compare the difference between the new UHERM model and the previous model from two aspects of availability and applicability. In terms of applicability, especially, we have increased the discussion on the mutual support of the new UHERM model and green infrastructure in detail.

Detailed remarks 1: L 27: introduction for acronym CA is missing 

Response: Accepted and revised.

In the revised manuscript, we added the full name of CA in the Abstract section.

Detailed remarks 2: L 116: you mention the cellular automata model here for the first time, then you also need to add “(CA)” to introduce this acronym 

Response: Accepted and revised.

In the revised manuscript, we added the acronym of CA in the introduction section.

Detailed remarks 3: L 131-132: wording needs improvement. eq. (3): should it be written C(t+a) = P x C(t) instead of having t+a in the exponent? 

Response: Accepted and revised.

We revised the Eq.3 according the suggestion.

Detailed remarks 4: L 221: the term heat disaster is undefined

Response: Accepted and revised.

In the revised manuscript, we refined the urban heat risk. The detailed modification process is described in the response of issue 2.

Detailed remarks 5: Section title 4.1: change “Based” to “based” 

Response: Accepted and revised.

We corrected the case error.

Detailed remarks 6: Table 4: which year is represented by columns and which one by rows? 

Response: Accepted and revised.

We have updated the header of Table 4.

Detailed remarks 7: L 335: check the wording of “disordered exploitation” 

Response: Accepted and revised.

We revised the wording of “disordered exploitation” to “overexploitation” in the revised manuscript.

Detailed remarks 8: L 336: change “transited” to “transitioned”

Response: Accepted and revised.

We revised the wording of “transited” to “transitioned” in the revised manuscript.

Round 3

Reviewer 3 Report

The manuscript has made major improvements including the English, which I’m quite happy about. Nevertheless, I still have some minor remarks (see below).

For the above mentioned reasons the manuscript is suitable for publication in Sustainability after minor revision.

Remarks:

Discussion:

In line 290 you write, “Compared with the existing urban heat environment prediction model, the new model has higher availability.” Which urban heat environment prediction models do you mean? And secondly, why does your model have a higher availability? This is also not what I meant initially. With availability, I mean whether or not your readers can use the UHERM model by contacting you or in some other way. This information should be added.

Detailed remarks:

l. 27: you could add the years in brackets after “initial year” and “middle year”

l. 47: change “result of primarily from urban developed land expanding to encroach upon” to “primarily resulting from expanding urban developed land which encroaches upon”

l. 76: change “environmental” to “environment”

l. 78: change “in mountain city” to “in Chongqing, which is a mountain city”. I think the name of the city or metropolitan area should be mentioned in the introduction

l. 140: no point after “and” needed

l. 177: I recommend to add Schwarz et al. 2012

Figure 3: Why is the medium temperature level of the process period associated with extreme risk and the higher temperature level with a lower risk class? This needs more explanation.

Table 3: Why is “” Highest temperature level” of the processing and prediction period not in the same row? Also, why are higher, medium and lower temperature levels each occurring in 2 risk classes? This needs more explanation.

l. 193: Maybe add a sentence to explain these stripes.

l. 251: and northern parts

l. 324: change “scientific” to “scientific basis”

References:

Schwarz, N., Schlink, U., Franck, U., & Großmann, K. (2012). Relationship of land surface and air temperatures and its implications for quantifying urban heat island indicators—An application for the city of Leipzig (Germany). Ecological Indicators, 18, 693-704.

Author Response

Part 1: Response to Reviewer #3 

Remarks: 1: In line 290 you write, “Compared with the existing urban heat environment prediction model, the new model has higher availability.” Which urban heat environment prediction models do you mean? And secondly, why does your model have a higher availability? This is also not what I meant initially. With availability, I mean whether or not your readers can use the UHERM model by contacting you or in some other way. This information should be added.

Response: Accepted and revised.

As we mentioned in the last revision, this new developed UHERM is based on CA-Markov model to simulate and predict urban heat environment risk. This UHERM avoids using absolute LST or geographical regression models which consider geographical factors, including population density, socioeconomic status with obvious regional development laws to predict the risk of urban heat environment. In the revised manuscript, we added some explanatory words for the existing model to make it clearer.

In the methodology section, we have introduced all the methods in detail. We think that the readers can use the data of their own research area to repeat the experiment according to the method. Of course, we also welcome readers who have difficulties or are interested in this research to contact us. The email address of the corresponding author has been listed in the article, and we also guarantee that this email address is available.

Detailed remarks 1: l. 27: you could add the years in brackets after “initial year” and “middle year” 

Response: Accepted and revised.

In the revised manuscript, we added the years in brackets after “initial year” and “middle year”. The year of 2005 is the initial year and 2010 is the middle year. Similar, we added the year of 2015 in bracket after “future year” in the abstract section.

Detailed remarks 2: l. 47: change “result of primarily from urban developed land expanding to encroach upon” to “primarily resulting from expanding urban developed land which encroaches upon”. 

Response: Accepted and revised.

In the revised manuscript, we have completed this revision.

Detailed remarks 3: l. 76: change “environmental” to “environment”. 

Response: Accepted and revised.

In the revised manuscript, we changed “environmental” to “environment” in the introduction section. Further, we searched the full text and corrected the same errors.

Detailed remarks 4: l. 78: change “in mountain city” to “in Chongqing, which is a mountain city”. I think the name of the city or metropolitan area should be mentioned in the introduction 

Response: Accepted and revised.

According to the reviewer's suggestions, we updated the introduction to the study case.

Detailed remarks 5: l. 140: no point after “and” needed 

Response: Accepted and revised.

In the revised manuscript, we delete the point after “and”.

Detailed remarks 6: l. 177: I recommend to add Schwarz et al. 2012 

Response: Accepted and revised.

We added the reference to the consistency of LST and air temperature.

Reference:

Schwarz, N.; Schlink, U.; Franck, U.; & Großmann, K. Relationship of land surface and air temperatures and its implications for quantifying urban heat island indicators—an application for the city of Leipzig (Germany). Ecol Indic. 2012, 18, 693-704, doi: 10.1016/j.ecolind.2012.01.001.

Detailed remarks 7: Figure 3: Why is the medium temperature level of the process period associated with extreme risk and the higher temperature level with a lower risk class? This needs more explanation. 

Response: Accepted and revised.

Upgrading from the medium temperature level in the processing period to higher temperature level in the prediction period, as a very typical trend of urban heat environment change, indicates that the region has a potential urban heat environment risk in the future year under the condition of continuous urban heat environment increasing. Therefore, this type of urban heat environment change is defined as extreme risk. In contrast, if a pixel both behaves as higher urban heat environment level in the processing period and the prediction period, it is defined as high risk area. Because, in this process, there is no obvious deterioration trend for the urban heat environment in the pixel, which shows that the current development and protection situation is acceptable. After that, if the protection level can be continued, the pixel may not show further warming trend as extreme risk zone.

Detailed remarks 4: Table 3: Why is “Highest temperature level” of the processing and prediction period not in the same row? Also, why are higher, medium and lower temperature levels each occurring in 2 risk classes? This needs more explanation.

Response: Accepted and revised.

First, the symbol “-” represents all levels of urban heat environment grades. In this statement, we can better understand the definition of urban heat environment risk between the processing period and the prediction period.

In theory, each pixel can represent any urban heat environment level in the processing period and the prediction period. Taking the highest temperature level for example, if a pixel is at the highest temperature level in the processing period, no matter what the urban heat environment level is in the prediction period, this pixel is defined as the extreme risk zone, although this possibility is very small. Because, in general, it is difficult to improve the highest temperature level to other temperature levels without taking the active projection measure. If a pixel is other temperature level in the processing period and the highest temperature level in the prediction period, then this pixel is also defined as the extreme risk zone. This phenomenon is very common and represents the deterioration of urban heat environment in the pixel. In additional, the combination rule was adopted for heat environment levels for the processing and prediction years. If a pixel met two risk level definitions at the same time, the higher risk level was selected.

Detailed remarks 5: l. 193: Maybe add a sentence to explain these stripes. 

Response: Accepted and revised.

In the revised manuscript, we added a sentence to explain these stripes. The stripes are related to the special mountainous terrain and rivers in Chongqing.

Detailed remarks 6: l. 251: and northern parts 

Response: Accepted and revised.

We corrected the previous statement.

Detailed remarks 7: l. 324: change “scientific” to “scientific basis”

Response: Accepted and revised.

We added “basis” after the word “scientific”.