1. Introduction
Natural disasters, particularly devastating weather-related disasters, cause significant concern worldwide. Approximately 80% of economic losses resulting from natural disasters are caused by extreme winds and associated events [
1,
2]. The World Meteorological Organization (WMO) reported that one weather, climate, or water-related disaster has occurred on average every day during the past 50 years, taking the lives of 115 people and causing USD 202 million in losses daily [
3]. In China, tornados are considered the most violent atmospheric hazard [
4]. They are a common weather phenomenon [
5], with approximately 108(±44) tornados every year, resulting in heavy losses to personnel, the economy, and the environment [
6]. Regarding spatial distribution, tornadoes in China mainly occur in the eastern plains, of which the middle and lower reaches of the Yangtze River are the frequent areas. According to statistical data from the National Climate Center from 1991 to 2020, China has an average of 38 tornadoes every year. Provinces such as Jiangsu, Guangdong, Hubei, and Anhui have a large number of tornadoes, with Jiangsu and Guangdong having the largest number. The annual average number of tornadoes is 4.8 and 4.3 respectively, followed by Hubei and Anhui, both of which have two. From the time distribution, the number of tornadoes from April to August accounted for 92% of the year, with July accounting for the highest proportion of about 30%. Judging from the occurrence time of strong tornadoes of EF2 and above, it also has the most in July [
7]. China began researching tornados in the 1980s, much later than in developed countries. Currently, scholars are exploring the climate mechanism of tornados and performing tornado monitoring on the regional to the national scale [
8]. Although the frequency of tornado occurrence has decreased consistently since the 1980s, several severe tornados have occurred in China in recent years [
9]. Due to the huge losses to the environment and society, significant efforts have been made to deal with these disasters. The losses to the ecosystems, economy, personnel, and buildings caused by wind-related disasters have been studied [
10,
11,
12]. Among all of these, the vulnerability of buildings is a crucial issue in the analysis of wind-induced physical damage [
13]. In particular, housing damage is a great concern due to casualties, resettlement, and post-disaster reconstruction [
14,
15]. However, information on the factors affecting housing damage is limited [
16,
17,
18]. Analyzing the factors influencing housing damage in tornados enables us to learn from disasters and reduce losses by improving disaster prevention measures.
A building damage survey can provide basic information on the disaster resilience of buildings and facilitate the analysis of building damage [
19,
20]. The building type has been used as a variable for evaluating hurricane-induced building damage [
21,
22]. Huo et al. [
23] investigated buildings damaged by Typhoon Soudelor in Chinese coastal areas, revealing common characteristics of the damaged houses such as inappropriate roof materials, aged structures, and the lack of integrity. Wind-induced housing damage includes exterior and interior damage. For low-rise buildings, the wind speed, exterior housing damage, and building shape have been used as variables for analyzing damage [
24]. The building structure is a key variable in damage investigations and vulnerability analysis. Light structures such as wood-frame houses are more likely to be damaged by hurricanes and earthquakes [
25]. The wind speed uncertainty is crucial in wind disaster analysis. In a tornado damage survey at Moore, Oklahoma, the wind speeds that damaged the residential structures were significantly lower than the established F-scale wind speeds, suggesting that low wind speeds during a tornado can cause severe building damage [
26]. Kwon et al. [
27] considered wind speed, building damping ratio, and terrain conditions in non-hurricane and hurricane winds. They found that the load factors were higher when the uncertainty in estimating the hurricane wind speed was accounted for [
28]. Morrison et al. [
29] investigated tornado-damaged houses in Vaughan, Ontario, and found that roof failure was due to internal pressurization and defects in the connections between the roof and the walls. Eight characteristics of buildings (roof, exterior wall, story number, year built, building area, floor, foundation, and shape) were used as indicators to establish a hurricane loss model. The higher the wind speed, the more building debris was produced, and the greater the external damage [
30]. In addition, the building height is also a crucial indicator of building safety during wind-induced events. Numerous studies have analyzed the response of tall buildings to wind events [
10,
12]. Hu et al. [
31] analyzed coastal topographic factors, community factors, buildings, and condition data on hurricane damage to residential structures. The results revealed that 1-story buildings were the most vulnerable to damage, but buildings with two or more stories were most likely to survive in a hurricane event. Pan et al. [
32] found that wind-induced internal and external pressures contributed to the net pressure in low-rise buildings with multiple openings. Failures of roofs and walls are common, especially in smaller residential buildings [
33].
Studies have shown that residential buildings in rural areas have less resilience and are easily damaged during severe natural hazards. Xie et al. [
34] investigated rural buildings damaged by a strong wind, revealing that rural houses were highly vulnerable to strong wind events. Zaini et al. [
35] found that rural houses were more vulnerable during windstorm events because they had lower engineering standards. In addition, kitchen houses, which are common in rural residential areas, significantly reduced the stability of the houses due to a height difference. Lam et al. [
36] analyzed the windstorm resilience of rural houses in the Temerloh district, Peninsular Malaysia. The study showed that high wind speed events caused severe damage to roof sheathings and non-structural components. Aquino et al. [
37] surveyed houses in two villages severely affected by the cyclone in Fiji, demonstrating building vulnerabilities caused by defects during the design and construction. Many physical measures to enhance building resilience to mitigate housing damage caused by hazards have been proposed [
38,
39]. Standohar-Alfano and van de Lindt [
40] performed a probabilistic tornado hazard analysis of residential wood-frame roofs. The results showed that stricter residential building codes were crucial and beneficial in regions with a high tornado risk. Ripberger et al. [
41] also proposed enhancing the building codes as a simple, inexpensive, yet highly efficient solution to reduce tornado-induced costs. Adelekan [
42] assessed 69 houses damaged by a windstorm event in Ibadan by interviewing residents to evaluate the houses’ vulnerability. A survey of the damage and structural failures after the tornado revealed that maintaining the integrity of the façade and roof systems could change the external conditions of critical structural components [
43]. Residential buildings face multiple hazards; thus, the same measure may have various effects on risk mitigation. For example, English et al. [
44] pointed out that the elevation of houses could increase their resilience to flooding and their vulnerability to wind, indicating that the higher a residential building, the more damage it might incur. However, mitigating the damage may significantly reduce the effects of hazards rather than improving the buildings’ resistance to violent tornados. Reducing unnecessary damage to structures in lower-intensity tornados such as EF0-EF3 tornados can also result in fewer losses [
45].
In summary, recent studies on housing damage during wind disasters have focused on building vulnerability to typhoons, hurricanes, tornados, and other events (e.g., the assessment of housing and environment damage, the structural performance of different building types, the influencing factors of losses, and the mitigation of the impact of wind disasters). Studies evaluating the influencing factors on housing damage have concentrated on the structural details of houses and wind speed. However, few studies have focused on examining environmental variables in disaster-hit locations, particularly in rural areas. The performance of different building types and structures in rural areas in China differs. Therefore, more research is needed. This study addresses the problem of disaster relief in rural areas, aiming to investigate the factors influencing housing damage during the Funing EF4 (
Table A1) tornado on 23 June 2016, considering the building condition, tornado intensity, and environmental factors. A field investigation was conducted to determine the damage index of the affected villages and the tornado damage pattern. The relationship between the building damage and the building conditions, tornado intensity, and environmental factors was analyzed using a multinomial logistic regression model and Pearson’s correlation analysis. The remainder of this paper is organized as follows.
Section 2 describes the study area and tornado and presents the field survey and data processing method.
Section 3 describes the results including the characteristics of the damaged houses, visualization of the damage index, and the tornado damage pattern. The factors influencing housing damage are described.
Section 4 provides the discussion and limitations of this study, and the conclusions and directions for future work are presented in
Section 5. The results provide information to reduce losses in wind disasters and support disaster relief and prevention planning in rural areas.
4. Discussion and Limitations
The mining of influencing factors of losses resulting from disasters is of great significance in reducing losses, improving resilience, and building a sound regional disaster prevention capability in future emergencies. In order to analyze the influencing factors of housing damage in relevant villages in the Funing tornado, three types of independent variables were considered including building conditions, tornado intensity, and village environmental factors to analyze the factors influencing housing damage in different villages caused by the Funing tornado. A damage index was established, and the damage pattern was examined to obtain a preliminary assessment of the tornado damage to houses. Subsequently, multinomial logistic regression analysis was used to determine the relationships between the degree of housing damage (SD, DR, and UR) and the independent variables. From the perspective of disaster relief, the findings of this study have practical significance for disaster relief and provide insights into reducing housing losses and casualties during wind disasters in rural areas.
The building area, building material, and the number of building stories were correlated with the housing damage level [
12]. Generally, the larger or taller a house, the more severe the damage. The post-disaster field investigation showed that some 2-story buildings were only slightly damaged, but most of these buildings were constructed in the past 10 years. Thus, their strong structural performance made them less vulnerable to tornados. In other words, the age of the house may also affect housing damage. This factor was not considered in this study. In addition, the building areas of rural houses in China are similar because the area is typically based on a standard, and expanding the BA is not permitted. The difference between the building areas is related to the ancillary facilities. A house with more ancillary facilities has a larger BA. Therefore, the impact of the BA on housing damage should be further explored. As previously mentioned, the construction age also affects housing damage. In general, the older the building, the lower its structural performance is in a disaster, which is true for tornados. Moreover, the building structure details (roofs, entrances, etc.), which were not covered here, were associated with housing damage in previous studies [
18,
29].
In the view of the impact of tornado intensity on house damage, the damage caused by high-intensity tornados is typically considered higher than that of low-intensity tornados for houses with the same area. In the current study, houses were more likely to be slightly damaged by EF4 tornados than by EF3 tornados, or by EF1 or EF2 tornados than by EF4 tornadoes. There are two explanations for this. First, there are differences in the building distribution between villages. If a building is very close to another, the wind speed may increase in strong wind conditions, causing the buildings to be damaged [
15]. Low-intensity tornados passing through areas with dense buildings cause more damage to houses than high-intensity tornados passing through areas with sparse buildings. In addition, the construction quality must also be considered. If the building quality of a village is low, low-intensity tornados can cause substantial damage, while the losses caused by high-intensity tornados may be the same. A study by Nateghia [
49] indicated that an F2 tornado was strong enough to destroy a wood-frame or nonreinforced masonry building. Most of the damaged houses were in the rural areas of Funing, and the low-quality houses could not withstand the tornado. Therefore, the correlation between the housing damage and the tornado intensity was weak for tornados stronger than EF2 because most houses in Funing were already severely damaged by the EF2 tornado.
The degree of damage to houses is, to some extent, related to the village environmental factors. In rural areas, trees are generally planted near the water, potentially protecting the houses. Moreover, the WAV in the village may have influenced the distribution and spatial pattern of the buildings, affecting the extent of tornado damage to the buildings. In addition, a large waterbody can affect the tornado strength and housing damage [
50]. Villages with a large VA typically have more buildings and a larger BAVB. The regression analysis showed that the number of houses with SD was associated with the total BA, which may be related to the path of the tornado (i.e., a low-intensity tornado passing through areas with high building densities). A village has a complex environment, and numerous factors should be considered such as topography, landform, residential settlement form, and green coverage. Since these factors may influence the housing damage in a tornado, they need to be explored in future studies.
Recent studies and our case study on the Funing tornado indicate many complex variables affecting housing damage in tornados, particularly in rural areas, where there are no fixed building codes, low structural performance of buildings, limited experience in constructing buildings resistant to wind disasters, and other risk factors. The current study analyzed the influencing factors of housing damage in a tornado using a multinomial logistic regression model. The results provide new information on reducing house losses in wind disasters in rural areas. However, knowledge of the relationship between housing damage and various dimensional factors remains limited, and this study has some limitations that may affect the results. In addition to the areas of improvement mentioned above, the classification of the houses by the investigators may have been subjective and may not reflect the actual level of destruction due to the limited investigation time and post-disaster reconstruction. Moreover, the data for this study were obtained from the post-disaster emergency survey, which had the primary goal of emergency response and the rapid resettlement of affected residents after the disaster. Therefore, there may be some limitations in the comprehensiveness of the data collection. In a wind disaster, housing damage is a result of a combination of man-made and natural factors. Therefore, to analyze the variables that affect the losses, it is necessary to consider the building conditions, disaster intensity, and external environmental factors as fully as possible. In any case, from the perspective of disaster relief practice in rural areas, this study is a reference in reducing losses when a potential disaster occurs.
5. Conclusions
Rural areas are vulnerable to disasters and tend to suffer more severe losses in destructive disasters than urban areas. An EF4 tornado occurred in Funing on 23 June 2016, killing 99 people, injuring at least 846 people, and destroying more than 2000 houses. We conducted a field investigation and analyzed the influences of the building condition, tornado intensity, and village environmental factors on housing damage using a multinomial logistic regression model and Pearson’s correlation analysis. The main findings of this study were as follows. (1) 2-story houses and masonry houses were more likely to have SD or be DR. The BA was positively correlated with houses in the SD and DR categories, indicating that the larger or taller a house, the more severe the damage. (2) Houses classified as SD were more likely to have been hit by the EF4 tornado than by the EF3 tornado, and the houses were damaged more by the EF1 or EF2 tornados than the EF4 tornado. This finding indicated that the level of housing damage was not strongly correlated with the tornado intensity in the Funing case, which may be related to the building density in the path of the tornados with different intensities. (3) Houses classified as SD had the highest correlation with environmental factors in the village. The number of damaged houses increased with the VA and BA. The proportion of houses with SD was positively correlated to the WAV, unlike the DR and UR houses. Moreover, the larger the WAV, the lower the housing damage.
The variables affecting housing damage in tornados are complex. This study took the rural areas in the Funing tornado as the study object to conduct a multinomial logistic regression analysis on the influencing factors of housing damage. Based on the results, the practical implications for improving the housing resilience of rural areas in wind disasters are as follows. First of all, building structures in wide rural areas should be enhanced, particularly for houses with individual kitchen and toilet areas; second, the spatial distribution of residential buildings in counties should follow a scientific basis for rural planning, as environmental factors are also associated with the housing damage in wind disasters; furthermore, as there are differences in building conditions, structure types, and disaster environments in rural areas across regions and ethnic groups, more research should be carried out to determine additional potential influencing factors on housing damage, thereby increasing the disaster resilience of rural areas.