Spatial and Temporal Distribution Characteristics of Active Fires in China Using Remotely Sensed Data

Round 1

Reviewer 1 Report (Previous Reviewer 2)

The reviewed version has taken into account my comments (at least the more important).

However, I would prefer, even after changes and clarifications, still a shorter section of results. Nevertheless, I understand authors want to present in an exhaustive manner their work and there is a source of results for a large and populated country that can interest readers.   

The discussion has been completely changed and now is appropriate and meets the requirements of a scientific paper.

After the major revision that the authors (with the addition of one new member) made, my overall opinion is that the paper can be accepted for publication.

Author Response

Point 1: However, I would prefer, even after changes and clarifications, still a shorter section of results.

Response 1: Thank you very much for your professional comments and guidance. As you said, the content of the results section is not concise enough. According to the content and structure of the article, we further optimized and condensed the section of results, especially the Section 4.5 of the results. At the same time, we added appropriate contents to the discussion section, making the section of discussion more integrated. In addition, we have modified the legend of Figure 16. Please see the revised version for details.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

Natural and/or man-made active fires are important drivers of the disturbance of global ecosystems. It is of great significance to reveal the spatial characteristics and dynamic changes of active fires for assessing regional biomass burning and carbon emissions. In this paper, the occurrence probability and intensity of active fires in Mainland China from 2001 to 2018 were analyzed to reveal the temporal and spatial distribution characteristics of active fires with different probability and intensity levels, providing a research basis for future fire prevention work. Overall, this manuscript is well written with an interesting topic. The structure and idea of the paper look appealing. The proposed technical framework is a well-established one and could be published by this journal with some improvement. For the benefit of the reader, a number of points need clarifying and certain statements require further justification. There are given below.

(1) It is suggest to display the source, web address, spatial resolution, main purpose, etc. of the data used in this paper in the form of a table.

(2) The conclusions obtained in the paper should be given necessary explanations, for example, what is the reason for the largest proportion of fire areas in Guangxi, Guangdong, Tianjin, Heilongjiang and other provinces.

(3) Why GWLR model is selected for fire risk assessment model instead of others needs to be explained.

(4) The contents of Section 4.5.4 Fire Risk Influencing Factors and Section 4.4 Active Fire and Fire Risk Influencing Factors seem to be duplicate. It is recommended to combine the above contents or further clarify their respective roles.

(5) The research conclusion of this paper should be compared with the relevant results of other scholars, especially the results obtained by using statistical data or other non remote sensing observation data.

(6) It is suggested that the research innovation or characteristics of this paper be clearly put forward in the discussion or conclusion section.

Author Response

Point 1: It is suggest to display the source, web address, spatial resolution, main purpose, etc. of the data used in this paper in the form of a table.

Response 1: Thank you for the suggestion. In the revised version, we listed the types, sources and purposes of the data used in this paper in the form of tables. The revised contents are as follows:

2. Data Sources

The data used herein mainly include five types, such as active fire, meteorology, DEM, NDVI, and socio-economic data (Table 1). Active fire data products mainly obtained from NASA's fire information resource management system (FIRMS). It has made certain progress in fire condition assessment [32]. This paper selected MODIS C6 active fire position vector data for subsequent analysis. The MODIS C6 data from 2001 to 2018 is mainly obtained by real-time observation of Terra and Aqua satellite sensors four times a day. Terra passes over the equator daily at approximately 10:30 AM and 10:30 PM (MLT) local time to obtain active fire information, with data acquisition starting in November 2000. Aqua (EOS PM) passes by the equator at about 1:30 PM and 1:30 AM (MLT) to obtain active fire information, with data acquisition starting in July 2002. The coverage is global, the spatial resolution is 1 km, and the coordinate system is WGS 84. Although there are VIIRS V1 data, which is mainly passed by the VIIRS sensor mounted on the S-NPP satellite near the equator, with a higher spatial resolution (375 m), it is shorter time series (2012 to date) that cannot meet the research needs [33,34]. Therefore, it was not considered.

In the research on the influencing factors of active fire in mainland China, many factors affect the occurrence and spread of active fire, and the relationship between each factor is complex. Referring to relevant studies [35–39], nine indicators including temperature (QW), precipitation (JS), surface temperature (LST), elevation (GC), slope (PD), normalized vegetation index (NDVI), distance to the nearest path (DL), population density (POP) and GDP were selected as fire risk impact factors from four aspects: meteorology, topography, vegetation, and human activities. The sources and main purposes of the above data, which all have the spatial resolution of 1 km, can be found in Table 1. The data processing are as follows: 1) The annual average temperature spatial interpolation data set, China Annual vegetation index (NDVI) spatial distribution data set, China GDP spatial distribution kilometer grid data set, and China population spatial distribution kilometer grid data set are used to calculate the average value of 2000–2018 through the grid calculator in ArcGIS, to obtain QW, GDP, pop factors. 2) The DL factor is generated by calculating the Euclidean distance from the road spatial distribution data. 3) The GC factor is obtained from the national DEM data, and then the PD factor is generated by the ArcGIS slope calculation. JS and LST factors are calculated by ArcGIS from the monthly precipitation dataset in China and the surface temperature dataset in China. Finally, the raster data of all factors were projected and reclassified. The projection is "China Lambert Conformal Conic".

Table 1. Data Descriptions.

Data	Data Sources	Purpose
MODIS C6	Fire Information for Resource Management System (https://firms.modaps.eosdis.nasa.gov)	Temporal and spatial distribution of active fire
Temperature	National Qinghai-Tibet Plateau Science Data Center (https://data.tpdc.ac.cn)	Meteorological factor
Precipitation
Surface temperature	Resource and Environmental Science Data Center of Chinese Academy of Sciences (http://www.resdc.cn)
DEM	Resource and Environmental Science Data Center of Chinese Academy of Sciences (http://www.resdc.cn)	Terrain factor
NDVI	Resource and Environmental Science Data Center of Chinese Academy of Sciences (http://www.resdc.cn)	Vegetation factor
Road vector	Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences (http://www.radi.ac.cn/)	Human activity factor
Population density	Resource and Environmental Science Data Center of Chinese Academy of Sciences (http://www.resdc.cn)
GDP

 

Point 2: The conclusions obtained in the paper should be given necessary explanations, for example, what is the reason for the largest proportion of fire areas in Guangxi, Guangdong, Tianjin, Heilongjiang and other provinces.

Response 2: We are grateful for the suggestion. The reasons for the largest proportion of fire areas in these four provinces are briefly mentioned in the Result section, and detailed explanations are given in the section of Discussion. At the same time, we have added two references to be pieces of evidence. The revised contents can be found in the revised version. The newly added references are as follows:

[57] Chang, Y.; Zhu, Z.; Bu, R.; Chen, H.; Feng, Y.; Li, Y.; Hu, Y.; Wang, Z. Predicting fire occurrence patterns with logistic regression in Heilongjiang Province, China. Landscape Ecol. 2013, 28, 10.

[59] Li, Y.D.; Feng, Z.K.; Chen, S.L.; Zhao, Z.Y.; Wang, F.G. Application of the Artificial Neural Network and Support Vector Machines in Forest Fire Prediction in the Guangxi Autonomous Region, China. Discrete Dyn. Nat. Soc. 2020, 2020, 5612650.

 

Point 3: Why GWLR model is selected for fire risk assessment model instead of others needs to be explained.

Response 3: Thank you very much for raising the question. Further explanation of the GWLR model will increase the readability of the article. Geographic weighted logistic regression (GWLR) is an extension of Binary Logistic Regression (BLR). It incorporates regression parameters into the geographic location of data, making regression parameters a function of the geographic location of observation points. OLR assumes that spatial variables are stationary variables, ignoring the spatial heterogeneity of model variables, and the fitting results of the model cannot fully reflect the spatial relationship of variables. GWLR takes into account the influence of geographical and spatial factors, as well as the local effects of spatial objects. It divides a large dataset into several small regions, reducing the differences between models, helping to improve model accuracy, and can be used to solve the problem of spatial stability [42]. In recent years, scholars have applied GWLR to fire prediction and spatial analysis of fire impact factors. Therefore, this paper chooses GWLR rather than other models. We also added detailed explanations in the section of Introduction and Discussion of the revised version. The reference is as follows:

[42] Wang, W.; Pan, J.; Feng, Y.; et al. Model and Zoning of Fire Risk in Gansu Province based on MODIS data and GODIS Imagery. Remote Sensing Technology and Application. 2017, 32, 514–523. (In Chinese)

 

Point 4: The contents of Section 4.5.4 Fire Risk Influencing Factors and Section 4.4 Active Fire and Fire Risk Influencing Factors seem to be duplicate. It is recommended to combine the above contents or further clarify their respective roles.

Response 4: Thank you very much for your excellent guidance. As you said, the contents of these two sections are a little repetitive. Section 4.4 mainly analyzes the relationship between active fire and fire risk factors. Section 4.5.4 mainly focuses on the spatial distribution of fire risk influencing factors and fire risk zoning. After comparing and further confirming the contents of the two sections, we have integrated and deleted the duplicate contents. In addition, some fire prevention suggestions in section 4.5.4 of the original manuscript are integrated into the discussion section of the revised version. In addition, we have modified the legend of Figure 16. The revised content is more concise and easy to read. The revised contents can be found in the revised version.

Point 5: The research conclusion of this paper should be compared with the relevant results of other scholars, especially the results obtained by using statistical data or other non remote sensing observation data.

Response 5: Thank you for helping us point out the deficiencies. We compare the results of this paper with those of other scholars, and the contents are present in the section of Discussion. The details are as follows:

Additionally, taking the Chinese mainland as the whole research area, the temporal and spatial variation characteristics of various fire types, the temporal and spatial distribution of occurrence probability and intensity, and the influence factors are analyzed and discussed from shallow to deep. As a developing country with the largest population in the world, it has a vast territory with complex and diverse landforms, climate types, and vegetation types. For instance, the climate is relatively dry in the north and wet in the south. The climate is dominated by the East Asian monsoon in East China, where the summer is wet and hot while winter is dry and cold [51]. It is known that about 75% of the land areas of China are covered by different vegetation types, mainly including forests, savannas, croplands, and grasslands [52]. Therefore, it is reasonable and appropriate to take the Chinese mainland as the research area. On the one hand, it can provide a reference for relevant international research. On the other hand, it is helpful for the planning and implementation of fire risk-related policies. In addition, most scholars devote themselves to the study of a certain type of fire. For instance, Ma et al. [53] used a random forest algorithm to analyze and study forest fires in China; Mohammadi et al. [54] took Iran as an example and studied the modeling of forest fire hazard areas based on the methods of Logistic regression and GIS. And more attention is paid to some areas with high fire incidences, such as Li et al. [55] analysis and research on the 16-year forest fire risk of typical subtropical regions in Zhejiang Province. Therefore, there may be some problems, such as insufficient attention to the overall situation of active fires and insufficient overall grasp. This study also shows and analyzes the typical provinces of active fire in the Chinese mainland (Heilongjiang, Guangxi, Henan, Shandong, Fujian, and Qinghai). The results showed that the dynamic changes in the frequency and intensity of active fires in the above provinces were similar to those in the Chinese mainland as a whole. At the same time, Zhang et al. [56] used MODISC6 data from 2001 to 2019 and VIIRSV1 data from 2012 to 2019 to analyze the spatiotemporal characteristics of active fires in the Arctic Region and establish a fire risk assessment model based on logistic regression. The results obtained from the two sets of active fire data are consistent, which further proves the scientificity of the research results in this paper. Chang et al. [57] used the data collected from the Chinese Forestry Science Data Center to study the forest fires in Heilongjiang Province from 1980 to 2009. Their research results are similar to the results of this study, which further explains the reliability of the results of this study. 

The newly added references are as follows:

[56] Zhang, Z.; Wang, L.; Xue, N.; Du, Z. Spatiotemporal Analysis of Active Fires in the Arctic Region during 2001–2019 and a Fire Risk Assessment Model. Fire 2021, 4, 57.

[57] Chang, Y.; Zhu, Z.; Bu, R.; Chen, H.; Feng, Y.; Li, Y.; Hu, Y.; Wang, Z. Predicting fire occurrence patterns with logistic regression in Heilongjiang Province, China. Landscape Ecol. 2013, 28, 10.

 

Point 6: It is suggested that the research innovation or characteristics of this paper be clearly put forward in the discussion or conclusion section.

Response 6: Thank you very much for your excellent guidance. There are four innovations in this paper. Firstly, this study presents the spatial distribution characteristics of active fires in the Chinese mainland from 2001 to 2018 at different time scales (such as annual scale and monthly scale). Secondly, taking the Chinese mainland as the whole research area, the spatio-temporal variation characteristics of various fire types, the temporal and spatial distribution of occurrence probability and intensity, and the influence factors are analyzed and discussed from shallow to deep. It is not only helpful to understand and grasp the spatio-temporal characteristics and development trend of active fires in the Chinese mainland, but also can provide scientific reference for the prevention of fire risks in countries or provinces as a whole and the formulation and realization of "double carbon" target planning. Thirdly, this study comprehensively explored the fire risk influencing factors from four aspects, including meteorology, terrain, vegetation and human activities. When considering the influencing factors of human activities, we take distance to the nearest path, population density, and GDP factors into account, which makes our research results more scientific. Then based on the GWLR fire risk assessment model, this study analyzed the fire risk influencing factors and their relationships in Chinese mainland, and drew the fire risk influencing factor zoning map. we put forward differentiated fire prevention suggestions for eight different types of fire risk impact factor zoning. In addition, it is further refined and summarized in t in the Discussion section. The details are as follows:

Firstly, this study presents the spatial distribution characteristics of active fires in the Chinese mainland from 2001 to 2018 at different time scales (such as annual scale and monthly scale). Secondly, taking the Chinese Mainland as the whole research area, the temporal-spatial variation characteristics of various fire types, the temporal and spatial distribution of occurrence probability and intensity are analyzed and discussed from shallow to deep. Thirdly, this study comprehensively explored the influencing factors of fire risk from four aspects: meteorology, topography, vegetation, and human activities. When considering the influencing factors of human activities, we take distance to the nearest path, population density, and GDP factors into account, which makes our research results more scientific. What's more, based on the GWLR fire risk assessment model, this paper analyzes the fire risk influencing factors and their relationship in the Chinese mainland, and drew the zoning map of fire risk influencing factors. We put forward differentiated fire prevention suggestions for eight different types of fire risk impact factor zoning. 

Author Response File: Author Response.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

General Comments

This manuscript aims to study the spatial and temporal distribution characteristics of active fires in China, using MODIS C6 data, together with spatial statistics. The objectives and the methodology are well justified. The article would be of interest to the audience of Fire.

Nevertheless, the manuscript needs an extensive revision of English for publication. There are several sentences that are not clear or confusing. The figures are mostly clear but there is a lack of information of the Chinese provinces. The discussion is currently poor and needs to be improved. However, I found several sentences in the Results chapter that can be moved to the Discussion.

I consider that this manuscript needs major revisions before is suitable for publication.

 

Specific Comments

Abstract, Lines 11-12: The “grade 1 and grade 2” should be replaced by “low intensity”. The readers don’t know yet what are “grade 1” and “grade 2”.

Lines 94-95: The authors should specify in which timezone are those hours (UTC?).

Figure 3: This Figure should also have the location of the Chinese provinces. If is not possible to name them all in a visible way, at least include the location of those who were selected.

Line 385: Correct the first word: Topographic…

Figure 15: Correct the legend: Extremely high fire risk area

Figure 16 d): This Figure is not visible. The authors should change the color or zoom in the area or areas of highest slopes.

Results: This chapter is too extensive and several sentences should move to the Discussion. For example, lines 558-563. The Results chapter should be only to describe the results and not to extensively analyze them (that part should be in Discussion).

Discussion: Currently, this chapter only discuss the negative parts of this manuscript. A complete discussion should focus on analyzing the results and explain also the achievements of this work. The authors should rewrite it.

Reviewer 2 Report

The manuscript presents a spatial description (grid 1 x 1 km) over two decades of active fires in China using remotely sensed data correlated to abiotic (e.g. temperature, precipitations) and biotic parameters (e.g distance from roads, population density).

According to authors the aim of the paper is to reveals the temporal and spatial distribution characteristics of active fires with different probability and intensity levels, providing a research basis for future fire prevention work.

The manuscript presents a low originality and don’t provide an advance in current knowledge in interpretation and correlation of factors correlated to active fires.

However, its exhaustive descriptive presentation could be more considered as a report than an original scientific manuscript.

There is a need this section to be shorter and more concise.

The discussion is extremely poor and inappropriate. There is no interpretation, analysis and explanation of the results.  The results are not put in the context of the overall research.  

There is a need to work hard on the discussion section.

 

My overall opinion is that the paper can be accepted for publication only after major changes.