1. Introduction
The heatwave is one of the main factors for weather-related illness and death across the world. The frequency, intensity, and duration of heatwaves have increased significantly due to climate change in recent years [
1,
2]. The elevated temperature ends with immense live loss, e.g., the Chicago heatwave in 1995 [
3], the European heatwave in 2003 [
4], and the Moscow heatwave in 2010 [
5] have led to more than 740, 70,000, and 10,000 deaths, respectively. Thirty percent of the global population is currently exposed to fatal climate conditions for at least twenty days a year. In addition, the threat to human life from extreme heat will increase if global warming continues [
6]. Extreme heat events are likely to cause severe human suffering and economic loss, therefore, increasing society’s resilience to these incidents is a critical challenge for government authorities and researchers.
Constructing spatial heat risk indicators is an effective method for quantitative assessment. Several frameworks have been developed to create heat risk indicators, while the frequently used methods are multiplication-based Crichton’s risk triangle framework [
7,
8,
9,
10,
11,
12] and summatory-based heat vulnerability index (HVI) framework [
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23]. Crichton’s risk assessment method has been used to assess flood hazards in the UK [
24] and climate-related heat risks [
25]. In addition, this method has been frequently applied to heat risk assessment in developing countries in recent years [
12,
22,
26]. The Crichton’s risk triangle framework states that heat risk is a function of hazard, exposure and vulnerability [
27,
28]: (1) heat hazard describes things that may cause a risk, which can be derived from the historical increasing trend of temperature, the current measured value or the predicted future value; (2) heat exposure commonly refers to the population exposed to heat environment, where the accurate population distribution is needed to calculate exposure to heat [
29]; (3) heat vulnerability denotes the vulnerable aspects of the exposed items to a given hazard [
28]. The heat risk index is derived from the multiplication of heat hazard, heat exposure and heat vulnerability [
9].
The HVI describes the structure of the physical environment and socio-economic conditions and how it responds to risks [
30,
31]. The approach has been adopted by the IPCC and applied widely, mainly in the country scale and cities with detailed census units and abundant statistical data. The HVI model also includes three aspects; exposure, sensitivity and adaptability. The content described by exposure is similar to the heat hazard index in the risk triangle framework, that is, the extent to which human is threatened by climate change [
30]. Sensitivity is also similar to high-temperature exposure; the extent to which people, natural assets or infrastructure respond to climate change and its effects [
30]. Adaptivity refers to the ability to reduce loss in the face of extreme temperature events, which is opposite to the heat vulnerability index [
30].
In recent years, the methods of heat assessment have been developed based on HVI and Crichton’s risk triangle framework. The common procedure is to select spatial variables that provide some indication of areas at higher or lower risk, followed by integrating these variables through either the unweighted or weighted approaches. The unweighted methods mostly refer to spatial overlay through the Geographical Information System (GIS) technique, while weighted approaches have been introduced in various studies. For example, Zhu [
22] used the analytic hierarchy process (AHP) and principal component analysis (PCA) to determine the weight of indicators for HVI. Rinner [
21] used the ordered weighted averaging (OWA) multi-criteria analysis to compose indicators for the heat vulnerability assessment of Toronto. In addition, Ho [
32] used multi-criteria analysis (MCA) to assign a weight to data layers when constructing heat vulnerability index for Canada. However, the majority of the heat assessment mapping studies utilized PCA or GIS analysis techniques to integrate parameters [
16], and few studies have compared the heat assessment mapping results of the GIS overlay and PCA method. Faisal [
33] assessed the outcome of urban environmental quality derived from GIS overlay and PCA, and pointed out the difference. According to their research, GIS overlay does not consider the correlation between variables, and the minimum number of components derived in PCA is indeterminable. These differences could cause variation in heat assessment mapping. Given that the recent heat risk assessment studies based on Crichton’s risk triangle framework used spatial overlay method [
8,
12] while the studies based on HVI framework frequently used PCA methods [
26,
34], and very few studies have compared the mapping difference between these two patterns; this study was designed to compare the spatial distribution of vulnerability from PCA method over the HVI framework with mapping results of heat risk from GIS overlay method over Crichton’s risk triangle framework.
The senior citizens are recognized to be more sensitive to heat than other populations [
31,
35,
36,
37,
38,
39,
40,
41], while children [
40,
42,
43,
44], the disabled [
26,
45,
46], and groups with low educational levels [
31,
45,
47,
48] are considered as the vulnerable groups in heat risk assessment. However, relatively less attention has been paid to the outdoor workers, who are constantly exposed to extreme heat environment and easily obtain work-related injuries [
49]. Impaired workers’ health can lead to a decline in human productivity and economic loss [
50]; adaptation and preventive measures are needed for these working groups. The literature has shown that apparent temperatures are reliable for assessing the working environment [
51], but most heat assessment research studies only consider air temperature and land surface temperature [
12,
17,
52]. In this situation, this study introduced wet bulb globe temperature (WBGT) as an indicator in heat risk assessment. It is a well-established heat index widely used in the assessment of heat stress where high WBGT has been proven to increase the risks of work injuries [
53]. WBGT is a combination of the natural wet bulb temperature, the black globe temperature and the air temperature; most formula calculating WBGT include meteorological variables related to temperature, humidity, wind speed and solar radiation. In this study, we calculated WBGT using maximum air temperature and relative humidity. Together with the number of hot days, land surface temperature (LST), and the air quality index (AQI) to provide a comprehensive description of the thermal environment that exposes heat stress on vulnerable population, including outdoor workers. In particular, the data were combined to depict a temporarily extreme heat environment that involves no temporal changes.
Previous studies have focused on the urban settings, while heat assessments in rural area are sparse and inconsistent [
54]. Sheridan [
53] and Wu [
54] have observed greater mortality and vulnerability in response to oppressive heat, in a rural rather than an urban location. At the same time, other studies indicate that urban residents are more vulnerable during heatwave [
55,
56]. These results suggest that the vulnerability to heat in the rural area is a multifaceted problem that involves factors such as public health infrastructure, heat risk awareness and sociodemographic characteristics [
57,
58]. However, the understanding of heat health of the rural population in different regions is still in an early stage, especially for developing countries that still have a large proportion of people residing in rural areas and are lack of a systematic geographical and census statistics. Therefore, this study using the northern area of Jiangxi Province in the developing country China as the study area to perform a regional scale research, exploring the high-temperature risk environment of both the urban and rural areas. Considering the data integrity and validity, the year 2015 was used as a bench mark for data collecting.
To summarize, the purposes of this study are: (1) to map heat risk environment using two methodologies developed over HVI and Crichton’s risk triangle frameworks; and (2) to compare the mapping results of two methodologies at the urban and rural areas.
Section 2 presents the material and methods; in which we give reasons why each variable was selected as heat risk environment indicator.
Section 3 presents the mapping results, as well as explicit explanation, followed by discussion in
Section 4, and a conclusion in
Section 5.
4. Discussion
This study presents a comparison of the heat health risk of the population in northern Jiangxi, China by spatial overlay and PCA methods, based on Crichton’s risk triangle and HVI framework. In Crichton’s risk triangle, as there is no commonly acknowledged standard weight for each parameter, all of them were equally weighted, and the area of highest heat risk was highlighted. In addition, the spatial distribution of heat risk is similar to that of the index with a great variation range, which is consistent with previous studies [
8]. PCA used in the HVI framework is more objective and can reduce the impact of population density with a large degree of dispersion. Although both of the methodologies pointed out the same highest risk area and are strongly correlated, heat risk index features relatively lower values than HVI within urban regions and a smaller variety of spatial distribution across urban regions.
The study area reflects the well documented urban heat island (UHI) phenomenon, which results in a conurbation being warmer than the surrounding rural areas [
85,
86]. As a planned economic development area, northern Jiangxi may continue to expand the city scale. Cities, as the economic and cultural focal point with high population density, are prone to severe environmental impacts, which accordingly means that they are particularly vulnerable to climate change [
87]. Both methodologies show the inner-city areas with high population density have high heat risk or HVI value, which is consistent with previous studies in the Yangtze river delta and Chongqing region, as well as in the USA [
8,
9,
31,
88]. Moreover, the spatial distribution of HVI shows that the non-central city area has a lower vulnerability; this could be attributed to better infrastructure and a relatively low proportion of the vulnerable population. Similar to other developing countries, there are densely populated areas outside the city which exhibits high heat risk [
89]. The highest heat risk and HVI value that occurred were observed in a suburban community in Poyang. Census statistics have shown that the abnormally high population density and relatively low level of economic development lead to this situation, which has been reported in multiple studies that increased heat risks are related to increased population density, both in urban and rural areas [
88,
90]. Other vulnerable counties were rural, clustered in the northeast of the study area, with living status, vulnerable population and road density as the primary drivers. This requires local residents to raise awareness of heat risk prevention and local government to strengthen the construction of public infrastructure, which is conducive to the prevention of heat hazards.
People engaged in a nature-based economy are closely associated with HVI. Meanwhile, although inaccuracy originates from the calculation of WBGT that wind speed in 2 m and solar radiation were not concerned, making the value of WBGT unable to represent the real outdoor thermal stress, the spatial distribution of the exposure index that is integrated with WBGT shows that the rural area exhibits much higher values. As many economic-economy and outdoor workers are working in rural areas and their intense physical labor, they are more likely to be threatened by severe heat stress. Xiang [
91] has found that excessive heat stress could lead to occupational heat-induced illness and an increase in medical costs and work days lost. Moreover, George Maier et al. [
18] have used AT as a variable to evaluate the vulnerability of Georgia, United States under heat stress based on the framework of HVI, and found that the death rate of heat stress weather was 13.4% higher than that of non-heat stress weather. Hence, WBGT explored in this research could serve as a good reminder that highlight the heat exposure in rural areas, as well as the heat-health related burden of the outdoor worker.
Land surface temperature or extreme temperature days show weak zero-order correlation with HVI. Actually, most of the fine scale raster-based data have relatively low zero-order correlation with HVI, which might be attributed to the differences in scale and resolution between the socioeconomic and environmental data. Despite this, the partial correlation suggested the strong influence on HVI from hot environment depicted by land surface temperature and extremely hot days. Moreover, as temperature extremes and variability will remain important determinants of health [
92,
93], spatial distribution of HVI emphasizes the area with high risk and reveals great variation across urban areas, and thus could provide suggestions for heat alerts and developing emergency interventions.
The strong correlation between HVI and heat risk index in most urban regions (8/9) proves that the mapping results of the two methodologies have good consistency in urban regions and can reflect the areas of high heat-health risk. It is predictable that the heat risk index values within urban regions are significantly lower than HVI values, as can be seen in the mapping result of the whole study area. Very few areas reached the high level of risk, and most of the heat risk index values are distributed in lower tail. In contrast, the spatial distribution of HVI could reflect the heterogeneity across counties, making it exhibit a larger variety across urban regions.
There are still some research gaps in this paper; (1) data of preexisting health concerns that denote vulnerability to heat conserved in the Chinese center for disease control and prevention, such as cardiovascular disease or psychiatric disorders, are not currently accessible [
26], adding difficulty to the verification of the results. Due to the different availability of data, the time range of variables was inconsistent, and the statistical indicators of each county in the study area were not unified, limiting the construction of indicators [
32,
94]. Some variables, such as home air conditioning, which is a strong protective factor against extreme heat events, have not been proxies in local statistics and are not considered in this research; (2) a typical period of high temperature is selected in this paper for the assessment of extreme temperature. Some studies have emphasized the spatial-temporal change assessment of heat risk in developing countries—this may be considered in future study—and (3) people’s ability to adapt to extreme heat environment will change, and extreme heat events may increase with global anthropogenic climate change. This involves many socio-economic, individual behavior and environmental change factors and needs further consideration [
95].
In the future, we hope to use a wider range of environmental and socio-economic data as well as remote sensing data of more explicit resolution, such as Sentinel or Ecosystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) thermal infrared sensor data to explore areas of high heat risk at local scale and during different time periods. Moreover, with fined-scale population statistics available, risk assessments aimed at particular categories of vulnerable population are in desideratum.