The Spatial Interface of Informal Settlements to Women’s Safety: A Human-Scale Measurement for the Largest Urban Village in Changsha, Hunan Province, China

Abstract

Building sustainable communities is always related to the occupants’ physical and psychological safety, environmental security and human settlements inclusivity, etc. The interfaces of Chinese urban villages have shown features that are irregular and chaotic, which led to potential safety hazards for women. This study revealed the blind spot of how environmental interfaces affect women’s safety and proposed interface micro-renewal strategies that would have sustainable positive impacts on the safety of vulnerable populations in the future. Taking the largest urban village in Changsha as a case study, a total of 764 samples were collected from human-scale surveys and interviews; t-tests and various regression models were adopted to explore the correlation of nine interface variables on their safety perception. Multiple equations were constructed through ANOVA analysis and the machine learning model (ROC). The results show that interfaces have a strong association with women’s perceived safety compared to men, especially women aged 41–55 who were not engaged and had manual work at night. The safety ranking of variables follows the penetrability > proximity > scale, but it shows differences in terms of work type and age. This study would provide the necessary research to complement a sustainable urban transition and gender equality in the informal settlements of the Global South.

1. Introduction

The Sustainable Development Goals emphasize the topics of gender equality and safe cities, in particular, the need for improving road safety in human settlements [1]. However, contemporary urban planning is mainly based on masculine demands [2] (pp. 229–256). Due to the ambiguity of land management, the informal settlements of global cities are usually characterized by dense and irregular forms [3], and urban villages (also called “village in the city”; “urban village” is abbreviated to “UV”) are defined as typical informal settlements [4,5,6]. In order to maximize the benefits of space, approximately 80–90% of the UV’s land has been built [7,8]. As a consequence, low-end living costs and cheap rental housing were rapidly created for millions of low-income migrants and flexible laborers [9,10].

Due to high mobility and excessive population density, the interface of urban villages is always in an active state of being transformed. Hence, the interfaces of the urban village (“interface of the urban village” is abbreviated as “IUV”), which is the intersection between spaces, systems or devices [11] (p. 66), are usually described as crowded, chaotic, sprawling, irregular, and complex fields [12,13]. These keywords are strongly associated with spatial crime and incidence [14] (p. 101), in particular, in the areas such as building-street edges, intersect corners, ground floor surfaces, etc. Crime prevention and sustainable development are closely linked [15]; the fear caused by low-quality living environments often leads to the exclusion of vulnerable residents from normal activities. Moreover, the vitality and economy of the community will also be severely affected.

The “Sustainable Urbanization in Asia” assigned by UN-Habitat claims that sustainable city services are not only about public health and economic development but also the safety of vulnerable groups, such as women and children [16] (p. 7). This study focuses on marginalized and vulnerable groups, namely women living in urban villages. The purpose of this study is to reveal the blind spot of how IUV affects women’s safety perception and behavioral patterns in order to renovate the interface variables based on the results. It aims to protect women, both physically and psychologically, from safety hazards in informal settlements and provide theoretical support and practical references for the renovation of old urban residential communities in the Global South, such as South China, India, Brazil, South Africa, etc.

1.1. Significance

If women’s safety demands are unsatisfied, fear will surface, which would continue to exacerbate the stigma of the UV environment; it would impact the sustainable development of the informal community, such as economic recession, fewer employment opportunities, low land value, and vitality dissipated, etc. Due to the large population and high rates of poverty in informal settlements, the quality of life in these areas has become a major concern [17], and living quality is considered one of the main indicators of sustainable urban development [18]. It is too late to take it seriously after the crime occurred because we cannot compensate for the loss and injury. Therefore, this is a significant topic for promoting sustainable urbanization, spatial justice and gender equality.

The interface of informal settlements (IUV) often involves power relations, particularly in contexts of urban inequality and marginalization. The IUV is significantly different compared with the interface of the normal residential community in the city, which mainly has shown in its morphological complexity and functional instability. Previous literature showed that IUV impacts residential security and safety perception [19], and women are more likely to be attacked and feel scared than men [20,21]. However, the previous literature paid insufficient attention to the impacts of the IUV on the safety of women’s activities. In this study, the unsafe factors of the IUV not only refer to the unfriendly scale for women at a physical level but also took into account the psychological differences in women’s sensations brought by IUV, such as fear, threat and pressure.

1.2. Innovation

Perceiving safety is a distinctive feature in human-scale spatial experience [22]. The human-scale measurement refers to exploring the environmental factors at eye level [23]; it aims at the ground floor level of the spatial interface that pedestrians can easily access, which means that the study is dedicated to paying attention to the influences of the microscale spatial variables on human safety perception. It has been proven that microscale spatial relationships have implications for urban street crime distribution [24], and urban design based on the human-scale perspective has shown significant social impacts on human safety [24]. Based on the “human priority” [25], the study changed the research perspective from the macroscopic “two-dimensional plane” to microscopic “three-dimensional space”. Researchers collected data through field surveys in the UV, combined with 382 photos of the interface’s activities from the perspective of human eyes, to conduct online surveys in order to provide a microscopic human-eye perspective on environmental research.

In China, urban informality is mainly manifested in the UV [26], and women always experience negative emotions during their mobility in such informal settlements [21]. From the perspective of urban planning, men dominate urban streets [27], and the patriarchal ideology results in women adapting their behaviors and hiding their vulnerable in public places [28]. Women have a higher rate of being hurt than men, such as physical aggression, intimidation and sexual harassment [29,30,31], and they prefer to choose to be silenced [32]. There are many studies that examined the safety factors of UVs [33,34,35]. Nevertheless, the differences in safety demands brought by gender preferences in the UV have still not been fully considered, and the present study is dedicated to finding the gap in the literature.

2. Conceptual Framework

The term IUV is more than building a façade physically, it could influence people’s activity schema and psychological perception as well. Based on the perspective of pedestrians’ eye level, Gehl categorizes interfaces from soft to hard edges, the soft edges with the contents of permeability, sociability and activity [36], and he clarified that the soft edge of the ground floor is beneficial to the safety perception in the street [37] (p. 99). Bobić classified urban interface morphology based on the range of urban complex phenomena [12] (pp. 15–18); he provides seven main types and forty sub-types of public/private interfaces [38]. Gehl’s classifications of interfaces discussed how urban interfaces could most effectively contribute to public space and urban life, while Bobić focuses on visual and psychological variables of interfaces and the space between interfaces.

Moreover, Kamalipour categorizes IUV as impermeable, accessible and porous [39]. Impermeable typically shows the characteristics of inactive and closed [40]. Accessible means up to 50% of the area of the interface is open to the public [20]. Porous shows the features of active and intense, which means more than 50% of the area of the interface is open to the public [41]. There are certain criteria related to the interface of informal settlements, including transparency, enclosure [42], directness, traffic modes [20], visibility [43], vitality [44] and distance [45]. The interfaces’ traffic modes, distance, and directness are closely aligned with the UV’s approach to regulation and laneways’ security; for example, the street does not allow vehicles to drive through because the distance of vertical interfaces is too narrow in the UV, it also means lack of vitality. Interfaces’ transparency is linked with visibility, which is also related to mutual surveillance. The enclosure of IUV, which is the height of interfaces on street sides, can give pedestrians a sense of openness, closeness, or squeeze, potentially affecting psychological safety.

Furthermore, Gehl points out that the ground-floor façade has the most significant influence on regional dynamics and activities [40] (pp. 101–108) because ground-floor interfaces can connect buildings and people [46] by accommodating a range of leisure activities, which might also contain many contradictions and conflicts as well [47] (p. 81). Interface at the ground floor level can determine the pedestrian’s walking experience [48], which contributes to the street’s vitality and dynamics by involving pedestrians’ behaviors with the public environment, such as visual connection, verbal interaction, social activities, etc. Moreover, the ground-floor interface can represent the perspective of the city’s eye level on the human scale [42] (p. 129); some vivid scenes of urban life are often displayed on the ground-floor interface, such as children playing football, street vendors hawking, fighting incidents, etc. Thus, it needs to be further emphasized that the study mainly focuses on the ground floor, which is accessible to pedestrians.

In this study, the authors constructed nine safety-related IUV variables in three main factors of penetrability, proximity and scale. These criteria of informal settlements’ interface determine whether pedestrians would effectively notice the potential hazards and whether it is conducive to pedestrians seeking help immediately in the face of danger.

2.1. Penetrability

Research has proved that the visual clearness of the street is more important to the perceived safety of women than men [49]. The penetrable interface can actively link the interior to the outside, which is easier for people to perceive what is happening, so the first main factor of IUV for women’s safety is penetrability. First of all, the concept of porosity plays a key role in understanding penetrability [39], and it refers to the structure permeability of the interface, such as whether there are windows or doors open to the public. The criteria range from porous to impermeable. If more than 50% of an interface is open to public space, it would be defined as porous [50], and vice versa. The number of porosities has a positive relation with the livelihood of the area [51], which benefits attracting pedestrian flows in the UV.

Secondly, transparency is another variable for penetrability, which is the material permeability of the interface. It plays an essential role in target hardening and natural surveillance [52]. Transparency can mediate the public gaze, achieving permeable communication at one extreme and privacy at the other [53], which enables visual exchange between the inside and outside [20]. It affects the penetration of pedestrians’ vision penetrability on the public side and is also related to the permeability of the human eye on the private side of the interface, and the criteria include semi-transparent, opaque and transparent.

Thirdly, due to the complexity of the IUV’s morphology, the interface on the ground floor is not usually aligned with the street [54,55]. Its concave and convex morphology would affect the penetration of sight [56]. Thus, flatness is the third variable for measuring line-of-sight penetration (Build-to-Line Ratio). The flatness refers to the degree of closeness of the interface from the public boundary, including the criteria of setback, set forward and alignment. This is the three-dimensional permeability of the interface related to sight penetration.

2.2. Proximity

The proximity of the interface refers to the distance and relationships [51]; it explains the relationship between interfaces and other security-related elements. The variables in the proximity are linked to urban security and human-scale measurement, so macro-planning variables are not included in the factor of proximity, such as greening, billboards, business types, etc. Continuity refers to the proximity of street shops; infrastructures point to the proximity between interface and types of secure facilities, and the cross angle aims to the degree of interfaces interacted, such as a blind corner is less than 90°.

The first variable of proximity is continuity. Women generally believe that hawkers, shopkeepers, young people, daily commuters and busy roads are safe, and the main reason that makes them feel unsafe is the gaze from strangers [57,58], and the more shops on the ground floor interface, the livelier it will be [42] (pp. 240–245). Hence, the high continuity of shops and stores in the UV would have positive effects on risk mitigation [59] (p. 194). The authors set up the criteria from Ashihara’s and Gehl’s works. The crowded Asian atmosphere exists in streets less than 20 to 25 m wide [60] (p. 49). Streets with 15–25 shops per 100 m have a higher vitality index, and 10–14 shops are pedestrian-friendly streets [42] (p. 241).

Moreover, the second variable of proximity is the distance between infrastructures and IUV, including types the monitor, signage and illumination. The “infrastructure violence” [61,62] can hurt women in many ways [63]; it concentrates on poor public facilities along with higher proportions of strangers [64], which is shown explicitly on lacking street signage, CCTV [65] and police booths [66], in particularly, poor lighting [67,68]. Hence, the completeness of the infrastructure on the IUV needs to be examined, including the indicators of surveillance, illumination and signage.

Furthermore, women around the world are more willing to take longer or different routes to avoid dead-end alleys, blind corners, and branch roads [69]. Therefore, it is necessary to pay attention to the cross angle where the two interfaces connect because it would create a large blind spot that potentially increases the sense of insecurity in women. The factors of visual penetration include visual distance, angle and field [70]. So, we use less than 90°, larger than 90° and no cross (straight) as criteria.

2.3. Scales

In addition to permeability and proximity, the scale of the interface itself is also closely related to safety. The scale of the spatial interface is closely related to the safety perception of pedestrians, including the ground, sides, and top. The scale of the ground surface implies the number of pedestrians that can pass together in public. The scale of the enclosure is visually defined by interfaces on both sides of the UV street; the larger the scale of the enclosure, the higher the sense of closeness. The scale of the sky exposure plane is a supplement to the scale of the enclosure; due to the ambiguity in spatial planning, factors such as wires, cantilevers, and thick trees will block the sky, thereby affecting the sight of pedestrians.

The first variable is the scale of the ground surface. Although the scale of the ground surface means the width of the street, it implies the number of pedestrians that can pass simultaneously in the UV. In the high permeable informal settlements, at least half to one meter of width is needed for solo pedestrians, and a lane width between two and three meters wide can prevent cars and parking lots from entering, which helps to create a vibrant vibe, and at least three meters width are usually kept to allow cars to access [71]. Therefore, the criteria for determining this variable depend on width, including width < 1.0 m (solo), 1.5~3.0 m (group) and >3.0 m (cars).

The second variable of scale is enclosure, which means the form of silhouettes that are visually defined by interfaces on both sides of the UV street [47]. This paper adopts the formula of street enclosure proposed by Ashihara as the variable criteria, which is the ratio of road width (D) to building height (H). People will have a sense of closeness when D/H < 1, and it will give a sense of symmetry when D/H = 1. The sense of distance would gradually show up (D/H > 1) as the ratio increases [71] (pp. 46–47).

Thirdly, women are known to be more sensitive to negative visual qualities [72], and the openness of the interface will affect the level of visual quality [73]. Since the authors use the width-to-height ratio formula (D/H) to calculate the interface enclosure, considering the vertical flatness of IUV would have the possibility of outward expansion by the cantilevers [63], wires and canopy [74]. It will also affect the quality of visibility and clearness. Hence, the author uses the proportion of the sky exposure plane as the third variable of the scale in order to compensate for the instability of the enclosure data; the sky exposure plane has three criteria, including narrow (open sky < 40%), medium (40~80%) and broad (open sky > 80%).

It should be further clarified that the purpose of this study is not to define the properties of the IUV but to analyze the weight of safety impacts of IUV variables on women through the framework shown in

Table 1. The index of interface variables of urban village.

Factor	Variable	Abbr.	Code	Criteria	Literature Review
Penetrability	Porosity	PO	PO1	open to public < 20%	Kamalipour [39]; Gao et al. [51]; Kamalipour [40]
			PO2	open to public 20~50%
			PO3	open to public > 50%
	Transparency	TR	TR1	opaque	Jacobs [52]; Bobić [11]; Dovey and Wood [19,43]
			TR2	Semi-transparent
			TR3	transparent
	Flatness	FL	FL1	setback	Oliveira et al. [56]; Jones [54]; Van Oostrum [55]; Kamalipour [40]
			FL2	align to the street
			FL3	set forward
Proximity	Continuity	CO	CO1	<15 shops/100 m	Ashihara [60]; Gehl [37]; Roy and Bailey [57]; Rišová and Madajová [58]
			CO2	15~20 shops/100 m
			CO3	>20 shops/100 m
	Infrastructure	IN	IN1	No indicator	Mahadevia and Lathia [75]; Datta and Ahmed [62]; Sadeghi and Jangjoo [67]
			IN2	1~2 indicators
			IN3	have all
	Cross Angle	CR	CR1	angle < 90°	Lin et al. [70]; UN-Women [69]
			CR2	angle > 90°
			CR3	straight
Scale	Ground Surface	GR	GR1	width < 1.0 m	Ashihara [60]; Dovey et al. [71]
			GR2	width ≈ 1.5~3.0 m
			GR3	width > 3.0 m
	Enclosure	EN	EN1	D/H < 1	Ashihara [60]; Ewing et al. [42]
			EN2	D/H ≈ 1
			EN3	D/H > 1
	Sky Exposure	SK	SK1	open sky < 40%	Tang and Long [73]; Van Oostrum [55]; Mundher et al. [72]
			SK2	open sky ≈ 40~80%
			SK3	open sky > 80%

and Figure 1 so as to adapt and transform it in future practice.

3. Materials and Methods

3.1. Case Study

Changsha is the capital city of Hunan Province, China, located in the lower reaches of the Xiangjiang River. It is an important transportation hub and one of the mega-cities in South China. The site is located in the Kaifu district, which is the central area of Changsha. It is the largest UV in Changsha, composed of Jixiang Lane and Ruyi Street; it covers an area of about 270,000 square meters and has a permanent population of about 21,000. Fourteen thousand square meters have been rented out, and a commercial area of 7840 square meters is still waiting for investment. Residents are plagued by problems such as aging houses, high crime rates, insufficient living facilities, etc. Although the local government provided appropriate subsidies to encourage renovation, it is a typical historical UV in Changsha. Today, the UV retains 13 immovable cultural relics, such as the Tongrenli Mansions, the first Western-style hotel in Changsha built in the 24th year of Guangxu in the Qing Dynasty (1898). Thus, it is not in line with its actual situation to improve the environment through large-scale demolition and reconstruction.

This study selected the road section with the highest commuting frequency in urban villages to conduct research. Since its layout has a strong linear distribution feature (Figure 2), the author divides the linear street into four sections (S1, S2, S3, S4) and codes each section into three spots, including head, middle, and end of the road from east to west (e.g., S1-1, S1-2, S1-3). So, it has a total of 12 spots in four sections (Figure 3). The research group shot 382 photos from 12 spots, and 120 photos were finally selected and shuffled for distributing the survey online. Considering that the participants could not patiently rate all 120 photos at one time, we divided these photos into five different sub-questionnaires that embodied 24 photos corresponding to interfaces at 12 spots in different time periods.

3.2. Research Design

There are two research questions in this study, and the first author designed the plans and objectives for both. The IBM SPSS Statistics 27 was used for data analysis, and GraphPad Prism 9.5 was applied for visualization. Besides the offline survey, all of the online survey was distributed through the platform Tencent Questionnaire.

The first research question is an online survey questionnaire that was conducted from 2 March to 12 April 2023, aiming to screen out the different safety demands of women compared with men in the UV in order to provide precise guidance for the question setting of the second research question. The second research question is offline interviews in the field through questionnaires, and it was conducted from 25 April to 18 May 2023. Based on the first research result, we focused on women between the ages of 18–55 and acquired the safety score by the Likert scale. The answering time is five to ten minutes, and every participant would receive a 5.00 ¥ cash reward.

There are four street sections in total, so the first author organized eight researchers into four groups (2 people/group); each group collected data from one section, one was responsible for documentary and shooting, and the other had to explain the question and to ensure the participants understood it.

Due to the population density [76,77], visibility [78,79] and weather [80] would have impacts on safety perception, so the research group chose sunny days with good visibility, moderate temperature and no strong wind to enter the UV in order to avoid excessive interference from other factors. Moreover, the levels of security are different due to the time periods [75], so the research group collected data both day (8:00–17:00) and night (18:00–23:00) on weekdays, rest days, and holidays.

3.3. Methodologies

The study used both quantitative and qualitative methods to analyze the data. The authors adopted Cronbach’s Alpha and KMO values to test the reliability of the two surveys. The independent-sample t-test (n = 359) was used to analyze the differences between women and men, and used the Pearson analysis to test the correlation of interface variables. It aims to verify the hypothesis that the urban village interfaces create more safety impacts on women than men and to find out the gender differences of insecurity concerns in the UV so as to provide guidance for the field survey.

In the field survey research, the author selected 12 representative spots (12 spots * 5 angles) in the UV plan drawn by AUTO CAD and asked female pedestrians to score them through interviews. Based on the scores, the safety heat maps of the urban village during the day and night were drawn graphically (n = 374). Then, the relationship between the interface and women’s safety perception is analyzed through certain model calculations, including simple linear regression analysis, multiple linear regression analysis, logistic regression, step-wise regression analysis, one-way ANOVA, and the receiver operating characteristic curve.

4. Results

4.1. Overview of Findings

4.1.1. Online Survey

The Cronbach’s alpha is 0.712 (α > 0.70), and the validity index KMO value is 0.716, which indicates the results are reliable (

Table 2. Reliability Statistics and Hotelling’s T-Squared Test of first online survey.

N of Items	Sample Size	Cronbach’s Alpha	Hotelling’s T-Squared	F	df1	df2	Sig
29	359	0.712	328,109.593	10,834.425	28	331	0.000 ***

*** p < 0.001.

and

Table 3. KMO and Bartlett’s Test of first online survey.

Kaiser-Meyer-Olkin Measure of Sampling Adequacy.		0.716
Bartlett’s Test of Sphericity	Approx. Chi-Square	1937.056
	df	231
	Sig.	0.000 ***

*** p < 0.001.

). A total of 370 samples were collected, and 11 of them were invalid; the effective rate was 97.02%.

The independent samples t-test was used to compare the different safety demands and perceptions of gender (

Table 4. Two independent-sample t-test on gender differences (n = 359) in security perception of interfaces.

		Mean ± Std. Dev
Factors		Male (n = 179)	Female (n = 180)	t	p-Value
Interface variables	Porosity	6.229 ± 1.931	6.133 ± 1.865	0.478	0.633
	Transparency	6.520 ± 2.254	5.706 ± 2.554	3.202 **	0.001
	Flatness	5.067 ± 1.880	4.583 ± 2.014	2.352 *	0.019
	Continuity	3.687 ± 2.239	3.872 ± 2.270	−0.778	0.437
	Infrastructure	5.927 ± 1.969	6.567 ± 1.986	−3.062 **	0.002
	Cross angle	7.061 ± 1.774	7.422 ± 1.654	−1.993 *	0.047
	Ground surface	4.911 ± 2.113	4.989 ± 2.003	−0.360	0.719
	Enclosure	2.939 ± 2.104	3.128 ± 2.028	−0.868	0.386
	Sky exposure	2.659 ± 2.560	2.600 ± 2.322	0.230	0.819
Worried	Sexual harassment	0.402 ± 0.492	0.622 ± 0.486	−4.262 ***	0.000
	Assault	0.492 ± 0.501	0.622 ± 0.486	−2.505 *	0.013
	Rape	0.173 ± 0.379	0.378 ± 0.486	−4.446 ***	0.000
	Stalking	0.581 ± 0.495	0.839 ± 0.369	−5.597 ***	0.000
	Robbery	0.469 ± 0.500	0.606 ± 0.490	−2.607 *	0.010
	Kidnapping	0.128 ± 0.336	0.333 ± 0.473	−4.736 ***	0.000
	Murder	0.112 ± 0.316	0.194 ± 0.397	−2.185 *	0.030
	Steal	0.559 ± 0.498	0.550 ± 0.499	0.165	0.869
	Drug dealing	0.196 ± 0.398	0.189 ± 0.393	0.159	0.874
	Do not worry	0.117 ± 0.323	0.033 ± 0.180	3.047 **	0.002
	Others	0.056 ± 0.230	0.072 ± 0.260	−0.631	0.528
Insecurity concerns	Psychological	0.648 ± 0.479	0.872 ± 0.335	−5.143 ***	0.000
	Physical	0.285 ± 0.453	0.500 ± 0.501	−4.265 ***	0.000
	No feeling	0.218 ± 0.414	0.056 ± 0.230	4.597 ***	0.000

* p < 0.05, ** p < 0.01, *** p < 0.001.

). Moreover, 179 males and 180 females participated in the online survey. The number of incidents that women worry about in the urban village is not more than men, but women worry more intensely because of the spatial interfaces; in particular, they are apprehensive about stalking (0.839 ± 0.369), sexual harassment (0.622 ± 0.486), rape (0.378 ± 0.486), kidnapping (0.333 ± 0.473) (Figure 4a), which is significantly different from males (p < 0.001).

Moreover, the means of “not worrying anything” in men is about 3.5 times higher than in women (male: 0.117; female: 0.033, p < 0.01). Although men are also most worried about being stalked, the mean value is still much lower than women. The incident of stealing ranked second in men’s choice (Figure 4a), which indicates that men are more inclined to worry about property safety in the UV, while women are mainly worried about their personal safety in the UV. Although both men (0.648 ± 0.479) and women (0.872 ± 0.335) in the UV are more biased toward psychological fear, the standard deviation shows that women’s psychological insecurity is more concentrated and intense than men’s. At the same time, there are far more men than women who have completely no worries about the IUV (Figure 4b).

Furthermore, the importance order of IUV on safety perception was shown by assigning weights of 1–9 points to the ranking results from survey participants. Nine points mean the most important to their safety, sorting in descending order. The Pearson correlation analysis (

Table 5. The Pearson correlations analysis results of online surveys (n = 359) on the importance of IUV variables to pedestrians’ safety perception (definitions for the abbreviations are in Table 1).

	PO	TR	FL	CO	IN	CR	GR	EN	SK
PO	1.000
TR	0.296 **	1.000
FL	0.226 **	0.353 **	1.000
CO	−0.302 **	−0.303 **	−0.300 **	1.000
IN	−0.365 **	−0.457 **	−0.394 **	0.181 **	1.000
CR	0.050	0.088	0.107 *	−0.327 **	−0.118 *	1.000
GR	−0.320 **	−0.379 **	−0.320 **	0.099	0.107 *	−0.156 **	1.000
EN	−0.176 **	−0.311 **	−0.208 **	−0.132 *	−0.011	−0.167 **	−0.007	1.000
SK	−0.292 **	−0.338 **	−0.362 **	−0.033	0.073	−0.246 **	−0.021	0.022	1.000

* p < 0.05, ** p < 0.01.

) shows that spatial interface variables are related to each other, which is the most common statistical method for studying the relationship between two variables [81] (p. 75). The first online survey proved the differences in safety considerations brought by interfaces between men and women, which proved the necessity of the hypothesis.

The results show that women’s fear of crime is more obvious than men’s, and interfaces can serve as an important medium to enhance the spatial safety of urban villages. If we do not take the safety impact of IUV seriously, it will not be beneficial in achieving gender equality in the goal of sustainable development. Hence, the authors launched the second field survey on the basis of the first research results.

4.1.2. Field Surveys

A total of 394 samples were collected in the field survey research, and the effective rate was 94.92% excluding 20 invalid samples. The Cronbach’s alpha with 38 questions from 374 samples was 0.824 (α > 0.70), and the KMO value was 0.782, which means the data is highly reliable (

Table 6. Reliability Statistics and Hotelling’s T-Squared Test of second offline survey.

N of Items	Sample Size	Cronbach’s Alpha	Hotelling’s T-Squared	F	df1	df2	Sig
38	374	0.824	23,139.494	565.032	37	337	<0.001 ***

*** p < 0.001.

and

Table 7. KMO and Bartlett’s Test of second offline survey.

Kaiser-Meyer-Olkin Measure of Sampling Adequacy.		0.782
Bartlett’s Test of Sphericity	Approx. Chi-Square	4253.878
	df	703
	Sig.	0.000 ***

*** p < 0.001.

). Of the participants, 33.4% are manual workers, and 66.6% are not; 67.6% of them have a spouse, and 32.4% are single; 69% of them are locals, and others are migrants. Moreover, 44.4% are aged 18–30, 33.7% are aged 31–40, 20.1% are aged 41–55, and 1.9% are other ages.

The researchers selected five different view angles (A, B, C, D, E) in 12 spots and led female pedestrians to rate the safety levels of interfaces with scores of 1–5. The heat map graphically shows the safety levels of IUV during the day and night (Figure 5), the horizontal axis represents the location, and the vertical axis represents the angle of the interface. The color means that the bluer, the safer and the redder, the more dangerous to women. All the spots in Figure 5 can be found in corresponding positions in Figure 6. Overall, the safety score at night is averagely lower than daytime. Nevertheless, women feel safer at night in some spots, such as (A, S1-3), (D, S1-3), and (C, S3-3), which are all located at the end of each street section. Moreover, there are certain spots that showed low scores either in the daytime or at night, such as (C, S1-1), (C, S1-2), (D, S3-2), and (D, S4-3). In some spots with relatively high scores, the cross angles of (B, S2-1) and (A, S4-1) are all greater than 90°; IUV of (D, S3-3) has high transparency, lighting, and high continuity of shops.

Figure 6a presents that the safety score in the middle of each section is averagely lower than the head and end of the section. The head and end of each section are connected with the secondary branch road, and the traffic flow is relatively lower in the middle of the section. It potentially proves that the intersection of the secondary road and the main street makes women feel safer (two interfaces connected). Then, the reddest spots are usually accompanied by third-level branch roads. In Figure 6b,c, the place marked by the yellow dotted line is a third-level road, and the width of the ground surface is less than 2 m, which also means that the scale of the width of UV’s branch road would have an impact on women’s sense of security.

4.2. Model Calculation

4.2.1. The Safety Impacts of Interface Variables on Women

In order to dive deeper into the connection between IUV variables and women’s sense of security, female participants rated the importance of nine interface variables to their safety in the field research. Based on this data, this study constructed three regression equations corresponding to different time periods. Moreover, the dependent variable was estimated by using multiple independent variables so as to explain and predict the value of the interface variable. Randomization was used to control the potential confounding factors, and the multivariate analysis correction was applied to simultaneously analyze the impact of multiple confounding factors on the outcome according to the type of dependent variable.

Because there are nine independent variables on the spatial interface, the number of independent variables is more suitable to adopt the stepwise regression model to test the safety impacts of variables on women.

The equation formula choice [82]:

$$y\mathrm{n}=\beta 0+\beta 1xn1+\beta 2xn2+\cdot \cdot \cdot +\beta kxnk+\epsilon n$$

(1)

By adding the scores of each spot, three safety scores for the all-day, day, and night are correspondingly obtained. The safety score is the dependent variable (Y), and the scores of the nine spatial interfaces are the independent variable (X). There are nine independent variables and one dependent variable in each calculation, and three stepwise regressions are performed (

Table 8. Step-wise regression analysis (n = 374) of field surveys on interface variables in different time periods.

		Unstandardized Coefficients		Standardized Coefficients	t	Sig.	95.0% Confidence Interval for B	Collinearity Statistics
		B	Std. Error	Beta				Tolerance	VIF
a. Sum of safety score (all day)	(Constant)	67.441	1.065		63.319	<0.001	65.346~69.535
	Cross angle	6.694	1.364	0.247	4.907 ***	<0.001	4.012~9.377	0.978	1.023
	Enclosure	−4.339	1.397	−0.155	−3.105 **	0.002	−7.087~−1.591	0.987	1.013
	Sky exposure	3.076	1.244	0.124	2.473 *	0.014	0.631~5.522	0.988	1.012
b. Safety score in the daytime (8:00–17:00)	(Constant)	35.680	0.514		69.443	<0.001	34.67~36.691
	Transparency	1.660	0.808	0.107	2.054 *	0.041	0.070~3.250	0.962	1.039
	Enclosure	−1.823	0.814	−0.115	−2.240 *	0.026	−3.424~−0.222	0.977	1.023
	Cross angle	1.682	0.799	0.110	2.106 *	0.036	0.111~3.2530	0.959	1.043
c. Safety score at night (18:00–23:00)	(Constant)	30.747	0.711		43.248	<0.001	29.349~32.145
	Cross angle	4.864	0.733	0.327	6.635 ***	<0.001	3.422~6.305	0.961	1.041
	Sky exposure	2.485	0.663	0.182	3.747 ***	<0.001	1.181~3.788	0.986	1.014
	Enclosure	−2.547	0.752	−0.166	−3.385 **	0.001	−4.026~−1.067	0.966	1.036
	Continuity	1.505	0.684	0.108	2.199 *	0.029	0.159~2.850	0.965	1.036

Durbin-Watson Value: a: 1.933, b: 1.869, c: 1.969. * p < 0.05, ** p < 0.01, *** p < 0.001. Stepwise (Criteria: Probability-of-F-to-enter ≤ 0.050, Probability-of-F-to-remove ≥ 0.100).

).

The stepwise regression result indicates the temporal difference in the impact of the spatial interface on women’s safety perception (p < 0.05). The VIF has a positive relation to the collinearity, and the greater the VIF, the more severe reciprocal of tolerance [83]. There is no significant multi-collinearity in the models when the VIF < 10, while all of the variables’ VIF in Table 8 are less than 2. Based on step-wise regression coefficients, the larger the angle of interfaces, the more secure women feel. The larger the area of sky exposure, the safer women feel. The lower the ratio of enclosure (D/H < 1), the more secure women feel. The higher the continuity of the shop in the interface, the safer women feel. The equations are presented below:

$$\mathrm{All} \mathrm{day}=67.441+6.694\ast \mathrm{Cross} \mathrm{angle}-4.339\ast \mathrm{Enclosure}+3.076\ast \mathrm{Sky} \mathrm{exposure}$$

(2)

In all time periods, enclosure shows a negative effect on the degree of security, in which the higher the ratio of D/H, the lower security, and the effect is more significant at night than during the day. In all time periods, the cross angle has the most significant impact on women’s safety perception among these independent variables. In the all-day, the safety impact of IUV variables on women is cross angle > sky exposure. Once other variables remain unchanged, the overall security increases by 6.694 on average when the cross angle increases by one unit, and the overall safety increases by 3.076 on average when sky exposure increases by one unit.

$$\mathrm{Daytime}=35.680+1.660\ast \mathrm{Transparency}-1.82\ast \mathrm{Enclosure}+1.682\ast \mathrm{Cross} \mathrm{angle}$$

(3)

From 8 am to 5 pm, the safety impact of IUV variables on women is cross angle > transparency. Once other variables remain unchanged, the overall safety will increase by 1.682 on average when the cross angle increases by one unit, and safety increases by an average of 1.66 for every unit increase in transparency.

$$\mathrm{Night}=30.747+4.864\ast \mathrm{Cross} \mathrm{angle}+2.485\ast \mathrm{Sky} \mathrm{exposure}-2.547\ast \mathrm{Enclosure}+1.505\ast \mathrm{Continuity}$$

(4)

From 6 pm to 11 pm, the safety impact of IUV variables on women is cross angle > sky exposure > continuity. Once other variables remain unchanged, the overall safety degree increases by 4.864 on average for every unit increase of cross angle. The overall safety increased by 2.485 on average for each unit increase in sky exposure. The overall safety increased by an average of 1.505 for every unit increase in continuity.

Furthermore, in order to focus more precisely, the research group added different questions in the field survey, including age, living status, marriage status, and work type, so as to clarify the differences in women. The authors used multiple linear regression models to present the influences of interface variables on different statuses of women. Since the variables in the multiple linear regression model are mainly quantitative values, qualitative indicators need to be converted into quantitative values before they can be applied to the regression equation. Therefore, the authors set the categorical variable as a dummy variable before bringing it into the calculation so as to obtain the results of multiple linear regression analysis (

Table 9. Multiple linear regression analysis (n = 374) on work type and age.

Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.	95.0% Confidence Interval for B	Collinearity Statistics
		B	Std. Error	Beta				Tolerance	VIF
1	(Constant)	67.610	0.760		88.966 ***	<0.001	66.116~69.105
	manual	6.878	1.315	0.262	5.232 ***	<0.001	4.293~9.462	1.000	1.000
	none manual	0.000
2	(Constant)	63.871	1.420		44.987 ***	<0.001	61.079~66.662
	manual	7.283	1.315	0.277	5.538 ***	<0.001	4.697~9.869	0.979	1.021
	18–30	4.779	1.615	0.192	2.959 **	0.003	1.603~7.955	0.585	1.709
	31–40	4.403	1.685	0.168	2.613 **	0.009	1.090~7.716	0.594	1.683
	41–55	0.000

Durbin-Watson Value: 2.077. ** p < 0.01, *** p < 0.001.

) and collinearity diagnostics (

Table 10. Collinearity diagnostics result (n = 374) on work type and age.

Model		Eigenvalue	Condition Index	Variance Proportions
				(Constant)	Manual	18–30	31–40
1	1	1.578	1.000	0.21	0.21
	2	0.422	1.934	0.79	0.79
2	1	2.311	1.000	0.03	0.07	0.03	0.03
	2	1.013	1.510	0.00	0.01	0.16	0.18
	3	0.569	2.015	0.02	0.82	0.02	0.12
	4	0.107	4.640	0.95	0.10	0.78	0.67

).

The authors conducted multiple linear regression analysis three times by setting the independent variables as dummy variables individually, and the effective result was presented in Table 9. The results showed that spatial interfaces had no significant effect on living status and marital status (p > 0.05), so it was eliminated in the multiple regression model. There is no significant difference in the safety perception between women aged 18–30 and 31–40 (p > 0.05), so aged 41–55 was used as the comparative variable. Women aged 18–30 and 31–40 rated UV as safe more than women aged 41–55 (p < 0.01), which shows a significant difference compared with women aged 41–55.

Based on this result, we can draw the formula for the safety perception in the urban village and women’s statuses:

$$\mathrm{Safety}\left(\mathrm{age}\right)=63.87+4.779\ast \mathrm{age}(18~30)+4.403\ast \mathrm{age}(31~40)+\mathrm{age}(41~55)$$

(5)

$$\mathrm{Safety}\left(\mathrm{work}\right)=67.61+6.878\ast \mathrm{manual}+\mathrm{none} \mathrm{manual}$$

(6)

The results would provide references for the sustainable renewal of IUV. The prediction aims to help urban planners and designers to recognize the insecurity hazard brought by low-quality interfaces more accurately, and it will potentially contribute to community resilience.

4.2.2. Correlation of Nine Interface Variables

In addition to overall examining the impacts of interface variables on women’s perceived safety, the first author also looked at how each interface criterion impacts women’s safety, as well as the association of penetrability, proximity, and scale. The research group acquired 24 items of 12 interfaces that were scored by women in the daytime and night. The first author selected the mean value of 24 items’ sum score (Mean _Max = 5, Mean _Min = 1). Then, based on the mean value result, these samples were divided into two categories, including safe type (3.5 < mean value < 5) and unsafe type (1 < mean value < 2.5), and a total of 35 samples were obtained at the end. Later on, five experts, including architects, engineers, professors of spatial design, professors of urban planning, etc., were invited to evaluate the criteria levels of interface variables corresponding to 35 samples’ locations (see the criteria levels in Table 1). Ultimately, three different models were constructed based on this representative 35 samples and the experts’ evaluation.

First, one-way analysis of variance (ANOVA) was used to analyze the experts’ evaluating results and the mean value of safety scores. According to the p-value in the Homogeneity of variance test, Tamhane (p < 0.05) or Bonferroni (p > 0.05) was adopted to present the results of the Post Hoc Tests [84]. Moreover, we used Welch and Brown-Forsythe for robust tests of equality of means when the p-value in the test of normality is lower than 0.05 [85].

In terms of the penetrability of the interface (

Table 11. One-way analysis of variance (n = 35) on the penetrability of interfaces.

Penetrability	I	J	Mean Difference (I−J)	Std. Error	Sig.
Porosity	PO1	PO2	−0.333	0.217	0.394
	PO2	PO3	−0.785	0.320	0.086
	PO3	PO1	1.118	0.261	0.010 *
Transparency	TR1	TR2	−0.443	0.227	0.247
	TR2	TR3	−1.063	0.233	0.006 **
	TR3	TR1	1.506	0.112	0.000 ***
Flatness	FL1	FL2	0.710	0.221	0.011 *
	FL2	FL3	−0.052	0.136	0.975
	FL3	FL1	−0.658	0.191	0.010 *

* p < 0.05, ** p < 0.01, *** p < 0.001.

), only porosity > 50% shows a difference with porosity < 20%. Transparent presents significant differences with both opaque and semi-transparent, but opaque and semi-transparent do not show influences on women’s safety perception. In the variable of flatness, setback shows positive correction with women’s sense of safety.

In terms of the proximity of interface (

Table 12. One-way analysis of variance (n = 35) on the proximity of interfaces.

Proximity	I	J	Mean Difference (I−J)	Std. Error	Sig.
Continuity	CO1	CO2	−0.314	0.200	0.377
	CO2	CO3	−0.970	0.237	0.001 **
	CO3	CO1	1.284	0.175	0.000 ***
Infrastructure	IN1	IN2	−0.795	0.280	0.076
	IN2	IN3	−0.701	0.308	0.149
	IN3	IN1	1.495	0.159	0.009 **
Cross angle	CR1	CR2	−0.542	0.164	0.017 *
	CR2	CR3	−0.559	0.297	0.228
	CR3	CR1	1.101	0.252	0.009 **

* p < 0.05, ** p < 0.01, *** p < 0.001.

), there is no significant difference between <15 shops/100 m and 15~20 shops/100 m in continuity, but >20 shops/100 m shows differences with the other two criteria. The infrastructure shows that women’s perception of safety will be affected if they lack any indicator, including surveillance, illumination, and signage. There is no significant difference between a cross angle > 90 and a straight one, but it indicates that the larger the angle of the interface, the safer it is, and a straight one makes women feel safe the most.

In terms of the scale of interface (

Table 13. One-way analysis of variance (n = 35) on the scale of interfaces.

Scale	I	J	Mean Difference (I−J)	Std. Error	Sig.
Ground surface	GR1	GR2	−0.452	0.202	0.152
	GR2	GR3	−0.888	0.274	0.018 *
	GR3	GR1	1.340	0.193	0.001 **
Enclosure	EN1	EN2	−0.065	0.644	1.000
	EN2	EN3	0.821	0.252	0.023 *
	EN3	EN1	−0.757	0.602	0.807
Sky exposure	SK1	SK2	−0.585	0.228	0.045 *
	SK2	SK3	−0.181	0.283	1.000
	SK3	SK1	0.766	0.249	0.013 *

* p < 0.05, ** p < 0.01.

), the width of the ground surface > 3.0 m is most beneficial for women’s safety. The enclosure part shows that D/H ≈ 1 and D/H > 1 have significant differences in women’s safety perception; D/H ≈ 1 is the ideal situation. The open sky ≈ 40~80% shows differences with <40% and >80%, and it means that the higher degree of the sky exposure panel, the safer women would perceive it.

Second, based on the ANOVA test, the authors plotted the distribution graphs to show the relationships between the criteria of interface variables (see the level of criteria in Table 1 and Figure 1) and women’s safety perceptions during the day and night (Figure 7). The graph shows that the variables of porosity, transparency, continuity, infrastructure, and ground surface conform to the normal distribution characteristics in any time period. The positive correlation between most of the independent variables and women’s safety perception means that the larger the criteria of these variables are, the higher the safety score will be. For example, the larger the angle and transparency of the IUV, the better for women’s safety. Urban designers can take these results to renew the IUV, which is important to promote spatial equality and urban sustainable development for vulnerable groups in the context of urban design and safety planning.

However, certain variables show different correlations with safety. It makes women feel safer when the flatness shows the feature of setbacks. Women will feel safer when the cross angle of IUV is greater than 90° or straight during the day, but it tends to have less impact at night. The enclosure is negatively correlated with safety, which means the smaller the D/H during the day, the more will be recognized as safe, and D/H = 1 will be recognized as safe by women at night. Moreover, the sky exposure shows a positive correlation during the night. However, 40–80% sky exposure plane is the safest for women in the daytime.

Third, the authors adopted receiver operating characteristic (ROC) to draw the boundary line between safe type (3.5 < mean value < 5) and unsafe type (1 < mean value < 2.5) through the trend of FPR(1-specificity) and TPR (sensitivity). The purpose is to use evaluation indicators to reduce the interference caused by different test sets and to measure the performance of the model itself more objectively [86]. Before drawing the curve, the author used binary logistic regression to calculate the values of predicted probability of three sub-variable values of each interface factor and calculated three times in total (there are nine sub-variables in three factors). Then, they brought the three final interface factor values and safety classification into the model of classification calculation, and the ROC curve is presented in Figure 8.

Furthermore, the authors calculated the maximum Jordan index by the formula [87] (p. 546–560):

$$J_{\max }=\max t\left\{\mathrm{sensitivity}(t)+\mathrm{specificity}(t)-1\right\}$$

(7)

When the Youden index is the largest, the most appropriate test cut-off value can be calculated. The Youden index of penetrability is J _{penetrability} = 0.815, the FPR(1-specificity) is 0.185, the TPR (sensitivity) is 1.000, and AUC is 0.975 (95% CI: 0.930~1.000, p < 0.01). The Youden index of proximity is J _proximity = 0.926, the FPR(1-specificity) is 0.074, the TPR (sensitivity) is 1.000, and AUC is 0.970 (95% CI: 0.919~1.000, p < 0.01). The Youden index of scale is J _scale = 0.801, the FPR(1-specificity) is 0.074, the TPR (sensitivity) is 0.875, and AUC is 0.970 (95% CI: 0. 898~1.000, p < 0.01). To sum up, penetrability, proximity, and scale have impacts on women’s safety perception in the statistical sense, and the result shows (

Table 14. Area under the curve (n = 35).

Test Result Variable(s)	Area	Std. Error a	Asymptotic Sig. b	Asymptotic 95% Confidence Interval
				Lower	Upper
Penetrability	0.975	0.023	0.000 ***	0.930	1.000
Proximity	0.970	0.026	0.000 ***	0.919	1.000
Scale	0.961	0.032	0.000 ***	0.898	1.000

The test result variable(s): Predicted probability, predicted probability, predicted probability has at least one tie between the positive actual state group and the negative actual state group. Statistics may be biased. ^a. Under the nonparametric assumption. ^b. Null hypothesis: true area = 0.5. *** p < 0.001.

) that AUC (Penetrability) > AUC (Proximity) > AUC (Scale).

5. Discussion

The “Urban Village Master Plan (2019–2025)” issued by the government has stipulated that large-scale demolition and construction are not allowed in urban villages anymore. So, directly demolishing or rebuilding the interface is not suitable for the current situation of Chinese urban villages. According to the study results, the authors proposed certain interface micro-renewal strategies which are applicable to urban sustainability transitions. These strategies are not only possible with low capital investment but also would produce sustainable positive effects in the future.

First, since the results show that the safety score at the junction of the main road and the secondary road is generally higher, the author suggests that the material of store interfaces at the junction of the main road and the secondary road could be replaced with glass, in order to strengthen the penetration of sight lines.

Second, the results show that the larger the interface angle is, the safer on women. Hence, the reflective mirror could be placed at the turning corner where the angle of IUV is less than 90°. It can help women to be aware of the situation on the other side of the corner in advance. Moreover, to amplify the dim lighting at night, it is suggested to wrap aluminum-coated PET film around the reflective mirror.

Third, the flatness of the interface shows a negative correlation to women’s sense of security. Hence, fixed and folding chairs could be installed on the wall of the setback area to encourage a series of daily activities based on not affecting the normal flow of the street. It aims to encourage people, such as children, the elderly, sanitation workers and mobile vendors, to stay longer in public so as to reduce the chances of women walking alone and strengthen informal mutual surveillance.

To sum up, compared with the study on the typology [25,46] and morphology [48] of IUV [47], this study paid attention to the safety issues of the vulnerable groups in space and combined the physical variables of the interface with the safety demands of women. These demands are not limited to personal safety in terms of physical attributes [49] but also point to spatial psychological safety. Then, all of the research data and results in this study were acquired from the perspective of the human eye level, which is all about human-scale measurement. Compared with the large-scale demolition and reconstruction of the urban village from the macro perspective [88,89], the interface micro-renewal strategies proposed by the authors are closer to the actual safety needs of female pedestrians. It can produce safety implications under the low capital investment, which is conducive to the realization of sustainable urban planning in terms of safety.

6. Conclusions

A safe community for women is not just about staying away from violence but also includes women being able to move through public spaces without fear and enter public places with confidence. Women are one of the main vulnerable groups that need to be protected in the context of urban safety planning. The IUV variables considered for female pedestrians could potentially cover the safety demands of the majority population because women are relatively weaker than men in terms of strength, speed, psychology, and reaction when they encounter dangers.

The first online result shows that women are mainly worried about personal safety, and men are inclined to property safety in the urban village, and interfaces have a strong association with women’s perceived safety compared to men. The second field of research shows that time, age, and work type have influences on women’s safety perception. In all these nine variables, the cross angle, sky exposure, enclosure, continuity, and transparency have shown significant safety impacts on women, and the overall correlation of interface factor is penetrability > proximity > scale. One of the integral creations of sustainable development is to construct a safe physical environment, and crime prevention is the prerequisite for it because it directly relates to people’s quality of life [90,91,92]. Low-quality interface tends to cluster in informal settlements and low-income regions [93], in which vulnerable groups, such as women, have to bear poor living conditions. This study is dedicated to preventing harm and violence in informal settlements for women before danger occurs, which is beneficial for shaping an inclusive and safe city.

Nevertheless, it is necessary to further subdivide the differences into the attributes of women, such as the differences in safety needs in women with disabilities. Moreover, the different types of IUVs will be examined in further research. These limitations will be studied in a future plan.