Review Reports

Round 1

Reviewer 1 Report

The article is devoted to an interesting scientific direction, but in my opinion the topic has been left unexpended. The article has a number of comments.

1.      The set of chosen parameters of the model is not clear at all. The authors do not analyze the distribution of protected areas, ongoing fire prevention measures, landscape neighborhood, but at the same time they analyze such indicators as the number of secondary school students, the number of primary school students, gross domestic product (GDP). Such fire risk modeling parameters look very strange. One gets the impression that the authors see schoolchildren as the main causer of forest fires, which is very strange and contradictory.

2.      In general, the description of the research methodology is poorly done and does not give any understanding of the chosen parameters of the model, nor of the model itself. Apparently, the resulting model is empirical and cannot be extrapolated to other territories. In this case, I would like to see the research algorithm that the authors proposed for the possibility of its extrapolation.

3.      There is no comparison of territories in terms of natural parameters of fire risk. Perhaps they are the leaders in the formation of risk levels.

4.      The revealed influence of these parameters on the occurrence of fires looks unconvincing against the background of the fact that the authors themselves indicate the absence of relationships between the parameters. The GDP indicator itself is an integral indicator and can only indirectly judge the well-being of the population, which probably leads to less interest in the development of forest areas by the local population. In addition, there is no analysis of the availability of forests in these zones, a type of nature management that is very likely not oriented in cities with high GDP to the direct exploitation of forest resources.

5.      The authors themselves point out that the occurrence of forest fires is mainly the result of the complex action of three types of influencing factors: weather and climate, fuel and ignition source. At the same time, they introduce completely different parameters into the model, ignoring the named groups.

Author Response

Response to Reviewer 1 Comments

Point 1: The set of chosen parameters of the model is not clear at all. The authors do not analyze the distribution of protected areas, ongoing fire prevention measures, landscape neighborhood, but at the same time they analyze such indicators as the number of secondary school students, the number of primary school students, gross domestic product (GDP). Such fire risk modeling parameters look very strange. One gets the impression that the authors see schoolchildren as the main causer of forest fires, which is very strange and contradictory.

Response 1: The protected areas only accounts for a small part of the study area and are located in part of the three counties. And there is no statistical socio-economic data according to the protected area. These human data are collected in the corresponding three counties. Therefore, there is no analysis of the protected area. The fire prevention strategy of the whole area is to prohibit the use of fire. The requirements in the study area are the same and there is no spatial difference. As a natural factor, the adjacent landscape is not considered in this paper.The selection of parameters used in the model was based on the results of previous studies. Many studies have demonstrated a significant correlation between socioeconomic factors and fire occurrence using population density and GDP variables (Guo et al. 2016; Nunes et al. 2016). For example, Guo et al. (2016) found that population density and per capita gross domestic product (GDP) had a positive impact on fire occurrence. Kim et al. (2019) revealed that the spatial distribution of wildfires was concentrated in or around cities and that wildfire probabilities were strongly correlated with variables related to human activities. Also, Arganaraz et al. (2015) found an increase in fire activity with increasing population density and suggested that this may be associated with an increase in ignitions. Therefore, we have selected population density and GDP as our model variables.

In addition, according to the results of Grala et al. (2017), which analyzed the impact of human factors on wildfire occurrence, it was found that wildfires caused by children were likely to occur in densely populated areas. And, many studies have analyzed the causes of wildfire ignition and have concluded that fires caused by children are a significant factor in wildfire occurrence (Grala et al. 2017; Zhang et al. 2010). Meanwhile, the number of students is positively correlated with the population, so the number of students could also reflect the spatial distribution of the population. Therefore, we believed that the density of students might be a further understanding of population density in wildfire occurrence events, and depending on our results, it was found that the density of middle school students is one of the important factors influencing the fire occurrence density.

In order to more clearly express the reasons for the choice of parameters used by the model, we added more descriptions in the Materials and Methods as follows:” The variables used in the model in this study were selected based on the results of previous studies [43, 59, 60]. The probability of fire occurrence has a strong correlation with variables related to human activity and accessibility [61]. Therefore, we chose the density of the total population, gross domestic product (GDP), and agricultural GDP as human activity-related variables, and the density of the impervious areas and cropland areas as human accessibility-related variables. In addition, Grala et al. [62] found that wildfires caused by children were likely to occur in densely populated areas. Moreover, many studies have found that fires caused by children were an important source of wild-fire occurrence [62, 31]. Also, the number of students was positively correlated with the population size, thus the number of students could also reflect the spatial distribution of the total population. Therefore, we thought that the number of students might be a further understanding of population density in wildfire occurrence events, so we also chose the density of primary and middle school students as variables.” on Line 135-147 in the revised manuscript. 

Point 2: In general, the description of the research methodology is poorly done and does not give any understanding of the chosen parameters of the model, nor of the model itself. Apparently, the resulting model is empirical and cannot be extrapolated to other territories. In this case, I would like to see the research algorithm that the authors proposed for the possibility of its extrapolation.

Response 2: The hierarchical linear model (HLM) is a maturity model, which is a model based on existing software (HLM, version 7.0) for the analysis of different level factors. The model could be used in domains with nested data and has been widely used especially in the socioeconomic field (Diez-Roux 2000; Kuo et al. 2014; Yang et al. 2018). The objective of this paper was to demonstrate that multilevel methods have usefulness in wildfire studies, and therefore we did not investigate model algorithms. Moreover, the characteristics of wildfire occurrence and its impact factors are spatially heterogeneous, and the analysis method and model could be applied in different regions, but the results indicate region-specific characteristics. For more details on the specific algorithm of the hierarchical linear model, see Raudenbush and Bryk (2002). We added more details on the description of the hierarchical linear model in the Materials and Methods as follows:” The HLM is mainly used in socioeconomic research, such as public health research [63], education research [66], and hospitality research [67], etc. The model could analyze the impact of different levels/ degrees of nested factors/ data on the research objectives. For more details on the specific algorithm of the hierarchical linear model, see Raudenbush and Bryk [68].” on Line 173-177 in the revised manuscript. 

Point 3: There is no comparison of territories in terms of natural parameters of fire risk. Perhaps they are the leaders in the formation of risk levels.

Response 3: In general, climatic, vegetational, and topographical factors are more related to fire occurrence than human factors (Wu et al. 2014), which is the reason why the explanation of socioeconomic factors to fire occurrence is only more than 10% of our study. Even so, socioeconomic factors could determine many complex fire patterns (Costa et al. 2011), thus the influence of socioeconomic factors on fire occurrence is critically worth exploring. Therefore, this study only examined the effects of human-driven factors and did not consider natural-driven factors. If natural factors were added, the level divisions of natural factors (multiple natural units) were different from socioeconomic factors (multiple administrative units), and the number of levels, categories, and relationships between them were more complex. In this study, in order to illustrate the usefulness of the multilevel model in wildfire research, natural factors were not considered. 

Point 4: The revealed influence of these parameters on the occurrence of fires looks unconvincing against the background of the fact that the authors themselves indicate the absence of relationships between the parameters. The GDP indicator itself is an integral indicator and can only indirectly judge the well-being of the population, which probably leads to less interest in the development of forest areas by the local population. In addition, there is no analysis of the availability of forests in these zones, a type of nature management that is very likely not oriented in cities with high GDP to the direct exploitation of forest resources.

Response 4: This study focused on analyzing the effect of data with nested relationships at different levels on wildfire occurrence. We also analyzed the interaction between the parameters in different levels, which do not necessarily have interaction between the parameters from different levels, so we pointed out in the study that there was no relationship between some parameters. In addition, GDP is an important factor influencing fire occurrence has been proven by many works of literature (Arganaraz et al. 2015; Guo et al. 2016; Kim et al. 2019; Nunes et al. 2016), which is the reason why we choose GDP as one of the model parameters. Ideally, forest resource development could be analyzed through GDP, but in practice, the composition of GDP is complex and includes primary, secondary, and tertiary industries, and unfortunately, we cannot obtain data on the percentage of forest resource development in GDP. Moreover, it is difficult to distinguish the various components of GDP in relation to forest resource development. Therefore, the analysis was conducted using GDP which is a comprehensive indicator. 

Point 5: The authors themselves point out that the occurrence of forest fires is mainly the result of the complex action of three types of influencing factors: weather and climate, fuel and ignition source. At the same time, they introduce completely different parameters into the model, ignoring the named groups.

Response 5: The objective of this paper was to investigate the influence of socioeconomic factors (as human ignition sources) on fire occurrence and the usefulness of multilevel models in wildfire research. If other factors were added, the number of levels, categories, and relationships among all factors were more complex, it would not be easy to reveal the influence of socioeconomic factors on fire occurrence. Therefore, these natural factors were not considered in this paper.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper  "aims to explore the extent that spatial variation in fire occurrence  is caused by socioeconomic factors at the county level and the prefecture level". The manuscript in the current form looks very well prepared. The wildfires are among the main problems for many countries and for a long time researchers have been trying to model the occurence of these fires. At this point the research is very topical and useful not only for the Chinese regions, but for all the scienbtific workers who research this problem. There are some small issues, bit I considered them like very impottant such a good study to become a perfect one:

1. Please provide more literature revire in the section "2. Materials and methods", subsection "2.3 Multilevel model" about the implementation of such method in previous research and please clarify in brief the the advantages of this model against some other.

2. Provide the "Conclussion" section with the short explanation there about the contribution of the current paper to these types of the research - what is new? It is not clear from the manuscript.

2. Include in the "Discussion" Part exact recomendation s based on this model about the fire prevention policy. You shoud at leats give any advice to the Chinese authorities there. The paper aims to incestigate socio-economic factors for aspatial variation of wildfires not only testing the multilevel model in this case.

I wish authors good luck with the paper.

Author Response

Response to Reviewer 2 Comments

Point 1: Please provide more literature revire in the section "2. Materials and methods", subsection "2.3 Multilevel model" about the implementation of such method in previous research and please clarify in brief the the advantages of this model against some other.Response 1: We have provided more literature about the implementation of the hierarchical linear model in the revised manuscript and clarified in brief the advantages of this model. We added the descriptions in the Materials and Methods as follows:” The HLM is mainly used in socioeconomic research, such as public health research [63], education research [66], and hospitality research [67], etc. The model could analyze the impact of different levels/ degrees of nested factors/ data on the research objectives. For more details on the specific algorithm of the hierarchical linear model, see Raudenbush and Bryk [68].Currently, the analyses of influencing factors in wildfire research were always carried out spatially related to fire distribution patterns at different scales, and the analyses of nested relationships between different levels of influencing factors were rarely conducted, which made it difficult to further explore the causes of the aggregation characteristics of the spatial distribution of wildfire occurrence. Applying the HLM to study the relationship between different levels of socioeconomic factors and fire occurrence could improve the existing single-scale analysis, as well as, partially address the multi-scale and cross-scale driving effects from different levels of socioeconomic factors on the distribution patterns of wildfire occurrence. Researches applying multilevel models to study fire is fewer.” on Line 173-187 in the revised manuscript. Point 2: Provide the "Conclussion" section with the short explanation there about the contribution of the current paper to these types of the research - what is new? It is not clear from the manuscript.Response 2: We have provided more descriptions about the contribution of our study to the fire occurrence research, in order to more clearly express the importance of this study. We added a short explanation in the Conclusions as follows:” The HLM was applied to address the multi-scale and cross-scale driving effects of factors with nested characteristics on wildfire occurrence at different scales, improving the situation that existing studies are all single-scale analyses.” on Line 405-408 in the revised manuscript. Point 3: Include in the "Discussion" Part exact recomendations based on this model about the fire prevention policy. You shoud at leats give any advice to the Chinese authorities there. The paper aims to incestigate socio-economic factors for aspatial variation of wildfires not only testing the multilevel model in this case.Response 3: According to the comments of the academic editor, we have deleted the statements on policies and recommendations in the revised manuscript. 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The article can be accepted in form

Author Response

Thank you for your comments, which have provided valuable help in improving the writing and logical aspects of our manuscripts.