Research on Design of Underground Space for Refuge Based on Environmental Psychology and Virtual Reality

Abstract

Underground spaces hold significant potential for enhancing urban resilience against disasters, a key dimension of sustainable urban development. However, due to persistent associations of underground environments with negative psychological perceptions, these spaces—despite their superior protective advantages—are often overlooked as viable refuge options during emergencies. Guided by the theoretical framework of environmental psychology, this research focuses on underground parking garages in Hangzhou, China as its primary research object. The target participants are residents of Hangzhou aged 18–58 years (encompassing diverse occupations such as students, office workers, and service industry employees), who represent potential users of such spaces as refuges. To explore human behavioral patterns, psychological responses, and needs related to underground refuge spaces, we employed a two-phase methodology: first, a questionnaire survey to capture broader behavioral tendencies and subjective perceptions; complementing this, Virtual Reality (VR) experiments—a more immersive method—utilizing semantic analysis and the Likert scale to assess psychological indicators influenced by underground environments. The experimental data were analyzed via mean analysis, correlation analysis, and multiple linear regression analysis to identify the key environmental factors that influence psychological responses, as well as their optimal design parameters. These analyses reveal significant correlations between various environmental factors and psychological indicators. This research synthesizes individuals’ psychological tendencies in underground environments and proposes quantitative physical design guidelines to meet fundamental psychological needs. The findings provide theoretical and practical support for the design of underground space for refuge and the development of sustainable urban emergency shelter systems, thereby contributing to resilient and sustainable urban development.

1. Introduction

With the growth of the global economy and advancements in science and technology, the urbanization level in many countries and regions has significantly improved. Nevertheless, the processes of urbanization and sustainable development have imposed growing pressures on the aboveground built environment. Challenges such as population saturation, traffic congestion, environmental pollution, and resource shortages have become common issues faced by most cities [1,2,3]. These factors have led to an increase in the frequency of underground space utilization, which has gradually evolved into an additional spatial layer of the city [4]. In many cities, underground space development has achieved remarkable progress. For example, cities like Montreal and Tokyo utilize subway and pedestrian systems as the framework to connect public buildings, including stations and shopping malls, to form a systematic and extensive underground city [1,5,6]. European cities often utilize underground space to resolve urban land use conflicts and protect the urban environment, as well as natural and cultural landscapes, such as the Louvre in Paris and the Rijksmuseum in Amsterdam [7].

Underground space can not only meet the needs of urban expansion but also serve as supplementary disaster prevention resources in specific scenarios [8]. In China, many conventional underground spaces (including parking garages) incorporate basic civil air defense (CAD) design elements during construction, relating to the Code (GB50038-2005) [9]. This gives them a baseline level of structural stability, which can be leveraged for emergency refuge purposes. In terms of disaster resilience, underground spaces show advantages in certain scenarios: Their enclosed structure and soil covering effectively mitigate impacts from meteorological disasters (e.g., storms, extreme temperatures), as these hazards rarely cause direct structural damage to underground facilities [10]. They also provide relative safety from ground fires and small-scale industrial accidents, where the upper covering layer reduces exposure to external risks. The basic CAD features (such as reinforced doorframes and simple blast-resistant partitions) further enhance their ability to withstand minor explosive impacts or falling debris—common threats in urban emergencies. However, it is crucial to acknowledge their limitations: These conventional underground spaces, even with basic CAD elements, are not designed to resist high-intensity disasters. They cannot withstand large-scale wartime bombings, catastrophic earthquakes exceeding their design intensity, or direct hits from heavy weaponry. Their structural components (e.g., non-reinforced walls, ventilation systems) and surrounding geotechnical materials are vulnerable to damage under such extreme conditions, unlike specialized military or high-grade disaster shelters built with advanced reinforcement. Notably, historical examples—such as the adaptation of the London Underground as temporary refuges for civilians during World War I and II—highlight that underground spaces have long been repurposed to meet urgent shelter needs in emergency contexts. These cases underscore the adaptive potential of such spaces in civilian crises, though they are not intended to imply relevance to modern extreme warfare. Thus, the value of ordinary underground parking garages as refuges lies in their suitability for common urban emergencies (e.g., typhoons, chemical leaks, local disturbances) rather than extreme high-impact events. In recent years, as China strengthens urban emergency management, optimizing these existing spaces—with their baseline protective capabilities—into functional refuges has become a pragmatic approach to enhancing urban resilience [11].

However, underground space is often associated with negative associations and psychological concerns in people’s psychological perceptions, such as death, fear of confined spaces, and a sense of disorientation. People’s perception of the environment is mainly determined by their physical experience and impression of the environment [12]. Dong et al. [13] found that factors such as lighting, ventilation, and wayfinding systems in enclosed underground spaces directly influence human behavior and psychological states. Concurrently, Liska et al. [14] observed that surveillance blind spots and public risk zones in subterranean environments may lead to fear-driven avoidance behaviors. Fear and an unpleasant psychological perception of underground space may significantly reduce people’s willingness to enter such spaces during emergencies [15]. For example, during disasters, individuals often overlook underground refuge options despite their objective advantages in disaster prevention—a phenomenon linked to psychological barriers that the existing literature has inadequately addressed [16]. Notably, this points to a relevant observation: much of the existing discourse on underground shelter development tends to emphasize technical and functional aspects, such as structural safety and spatial capacity. In turn, this emphasis has led to a relative lack of empirical attention to how psychological perceptions mediate refuge behavior; there remains clear room to explore these understudied factors that shape how underground spaces are perceived and utilized. Concurrently, urban underground space utilization faces challenges in quality and system integration [17], partly because regulatory documents (e.g., GB50038-2005) only stipulate minimum hardware requirements, with no provisions for addressing higher-level psychological needs (e.g., reducing sense of enclosure, enhancing spatial comfort) [9,18]. This disconnect between technical standards and psychological demands hinders the effective utilization of underground spaces as refuges—an issue this study aims to address.

To address such psychological needs, environmental psychology provides principles that can be directly integrated into the spatial design of underground refuges, with a focus on shaping user behavior through targeted interventions. Kurt’s research on the interaction between people and their physical environment [19]—including his life–space theory and psychological field theory—highlights that spatial characteristics directly influence psychological states. This is supported by studies on architectural design and personal space [16,20,21], which demonstrate that physical features (e.g., lighting, openness) impact user comfort and willingness to engage with a space. Applied to underground refuges, this informs design strategies such as incorporating simulated natural light (e.g., daylight-mimicking panels in main corridors) to alleviate the “sense of enclosure”—a key barrier to utilization. By enhancing perceived openness, such design reduces avoidance behavior, making users more willing to enter during emergencies. Transactional theory in environmental psychology [22] further clarifies this dynamic: it posits that individuals and their physical environments mutually reshape each other, a process critical to underground refuge effectiveness. For example, designing intuitive wayfinding systems (e.g., consistent signage, color-coded zones) addresses the “fear of disorientation” in underground spaces. As observed in related studies [23], this not only improves spatial cognition but also fosters cooperative behavior by reducing anxiety, thereby enhancing collective refuge efficiency. These design interventions—rooted in psychological principles—ensure underground refuges meet not only structural standards but also psychological needs, directly addressing behavioral barriers to their utilization.

In recent years, Virtual Reality (VR) has emerged as a key visual research method at the intersection of architecture and environmental psychology, enabling high-accuracy simulation of 3D environments to assess visual impacts and psychological responses [24,25,26]. Related studies have used VR to explore path selection, virtual–real interaction, and spatial movement [27,28,29], though its application in underground refuge research remains limited. Furthermore, research within security studies employing Virtual Reality (VR) for psychological and behavioral assessment informs this study’s methodological approach [30]. Cocco et al. [31] implemented real-time VR simulations to measure affective responses and avoidance behaviors in controlled environments, specifically examining fear of crime relative to victimization risk.

Through a review of previous studies, it can be observed that, on the one hand, most existing research on the Design of Underground Space for Refuge (DUSR) remains at a qualitative level, with relatively limited quantitative investigation. For example, Lin et al. examined the indoor environmental conditions of confined sleeping areas in underground shelters during disasters by optimizing spatial design in terms of thermal comfort, air quality, and energy efficiency through environmental control systems [32]. Novalia et al. conducted disaster-prevention design analyses of underground buildings within transit-oriented developments (TODs) in several countries, exploring the adaptation of evacuation and refuge strategies to underground structures [33]. Bin et al. analyzed the types and characteristics of underground complexes with transportation hubs, integrating urban planning and disaster prevention design to propose evacuation systems and relevant design standards [34]. In China, civil air defense projects, which are commonly integrated into building basements, resemble the concept of disaster refuge spaces. However, the corresponding standards for civil air defense or refuge spaces remain minimal and largely theoretical, and there are few practical cases that can be identified as true implementations of DUSR. On the other hand, current applications of VR methods have primarily focused on topics such as way-finding, with very limited integration into DUSR research.

This research innovatively introduces environmental psychology and VR methods into DUSR, focusing on residential underground parking garages—an understudied refuge space. Unlike prior VR studies on aboveground structures, our VR simulations, paired with orthogonal design to control 11 physical variables, isolate their effects on psychological responses, integrating multi-disciplinary methods to determine optimal environmental parameters. Leveraging the visual and interactive capabilities of VR, we simulate real-world scenarios of these residential underground parking facilities (located in Hangzhou, Zhejiang Province, China) to analyze human psychological responses within such spaces and extract quantitative indicators. The objective is to identify key psychological impact factors associated with these underground refuge spaces during disasters, and to determine appropriate ranges for physical environmental parameters (e.g., lighting, signage) based on a thorough understanding of behavioral, psychological, and functional needs during evacuation. To achieve this, the research integrates psychological experimentation, statistical analysis methods, and architectural design principles. The conclusion data will provide a basis for the DUSR, support the formulation and improvement of relevant codes and standards, and contribute to the establishment of an urban emergency refuge system.

2. Materials and Methods

2.1. Research Procedure

The research process is as follows, shown in Figure 1: (1) Through questionnaire survey and analysis, behavioral patterns, psychological responses, and needs of people in underground refuge spaces are identified, as discussed in Section 3 (Questionnaire Survey Results). (2) VR experiments are then conducted to obtain psychological indicator data influenced by physical environmental factors in underground spaces, allowing for the identification of key physical factors affecting behavior and psychology, as well as the appropriate range for these physical indicators, as discussed in Section 4 (VR Experiment and Regression Analysis). (3) Based on the research content, strategies are proposed for constructing DUSR, as discussed in Section 5 (Design Strategies for DUSR).

2.2. Research Object

2.2.1. Types of Refuge Facilities

According to the Chinese national standard “Code for Design of Disasters Mitigation Emergency Congregate Shelter” (GB51143-2015, hereinafter referred to as “the Code”), refuge sites are primarily divided into emergency shelters and fixed shelters [18]. As detailed in

Table 1. Types of shelters and maximum opening time.

Applicable Places	Emergency Shelters		Fixed Shelters
Shelter Period	Emergency	Temporary	Short-Term	Medium-Term	Long-Term
Maximum opening time (d)	1	3	15	30	100

, the classification is based on two core criteria: (1) Functional positioning: Emergency shelters serve nearby populations for urgent or temporary refuge, while fixed shelters are equipped with accommodation and rescue-supporting facilities for centralized, long-term refuge. (2) Designed opening time: Emergency shelters are subdivided into “emergency” (≤1 day) and “temporary” (1–3 days) based on their response speed. Fixed shelters are categorized into “short-term” (3–15 days), “medium-term” (15–30 days), and “long-term” (30–100 days) according to their capacity for sustained refuge.

Against this classification framework, the selection of shelters in this study is logically derived: Focusing on disaster types for which underground spaces are suitable (e.g., sudden urban emergencies and prolonged public health events), we targeted emergency shelters (with maximum opening time of 1 day) and medium-to-long-term fixed shelters (30–100 days). This selection aligns with two key considerations: (1) Emergency shelters match the rapid response needs of short-term crises (e.g., terrorist attacks, sudden chemical leaks) where underground spaces can provide immediate refuge. (2) Medium-to-long-term fixed shelters correspond to scenarios requiring sustained refuge (e.g., large-scale public health emergencies), leveraging underground spaces’ enclosed structures for prolonged safety. By anchoring the selection to Table 1’s national standard classification, we ensure the research objects are both normatively compliant and tailored to the functional advantages of underground spaces.

2.2.2. Ideal Model Scenario

The main functions and contents of urban underground space include living space, transportation space, logistics space, commercial space, cultural activity space, production space, storage space, and disaster prevention space. Currently, the development and utilization of underground space in China primarily focus on transportation infrastructure. Taking Hangzhou as an example, more than 85% of the total underground space is used for parking [35], with most residential communities equipped with exclusive underground parking garages. These garages, as a form of static transportation infrastructure and independent underground structures, have standardized designs that meet national seismic and air defense requirements [36], while being widely familiar to local residents—factors that make them ideal for studying underground refuge adaptation.

Against this background, this research focuses on underground parking garages within residential communities in Hangzhou, Zhejiang Province, China—a site chosen for its representativeness: Hangzhou’s high proportion of underground parking spaces, coupled with residents’ daily familiarity with these structures, ensures the research findings can reflect common behavioral patterns and psychological responses in typical urban underground refuge scenarios.

2.3. Methods

2.3.1. Questionnaire

Referring to the existing literature, a questionnaire focusing on evacuation behavior, psychological responses, and needs in underground spaces was designed and distributed online. The contents of the questionnaire are shown in

Table A1. Questionnaire.

Question	Option
1. Your age?	□ Under 18 □ 18–28 □ 29–50 □ Over 50
2. Your gender?	□ Male □ Female
3. Your education level?	□ Junior high school and below □ Senior high school and junior college □ Undergraduate □ Undergraduate and above
4. Your industry?	□ Civil servant □ Company employee □ Worker □ Business □ Student □ Teacher □ Unemployed □ Other
5. Have you ever lived or worked underground?	□ Yes □ No
6. Have you ever experienced a disaster? (such as natural disasters, public health incidents, social security incidents, accidents and disasters, etc.)	□ Yes □ No
7. When you take refuge, your principle for choosing a shelter is?	□ Nearby □ Safe □ Follow the crowd □ Good environment □ Other (please specify) _____
8. Do you know the location of the emergency shelter announced by the city?	□ Yes □ No
9. When a disaster strikes, are you willing to take refuge in underground space?	□ Yes □ No
10. When what kind of disaster strikes, you are more willing to take refuge in underground space?	□Meteorological disaster □Earthquake □Flood □Nuclear accident □Communication and power supply failure □Epidemic □Terrorist attack □Fire □Other (please specify)______________________
11. When a disaster strikes, if there are clear signs to guide you, would you prefer to take refuge in underground space?	□ Yes □ No
12. If you stay in underground space for a medium- to long-term (about 30 days or more) to take refuge, what activities do you prefer to do during the “daytime” in the basement?	□Lying down to rest □Standing □Walking □Participating in public activities □Other (please specify)__________
13. What is your impression of underground space? (such as underground garage) (multiple choices are allowed)	□Enclosed □Safe □Fearful □Unsafe □Depressed □Strange □Dark □Other (supplement)________
14. During the medium- to long-term (about 30 days or more) ground refuge life, what aspects will you worry about? (Multiple choices are allowed)	□Eating □Drinking water □Electricity □Toilet □Washing □Getting information from the outside world □Mobile phone signal and Internet □Entertainment □Other (please specify)____________________________
15. If you stay in the underground space for a medium to long term (about 30 days or more), what aspects will you worry about? (Multiple choices are allowed)	□Eating □Drinking water □Toilet □Electricity □Washing □Getting information from the outside world □Mobile phone signal and Internet □Entertainment □Other (please specify)____________________________
16. Please choose D for this question	□A □B □C □D □E
17. When entering the basement for a long-term refuge (>30 days), what form of accommodation space can you accept?	□Personal private space □Open public space □Don’t care □Other (please add)________________
18. When entering the basement for a long-term refuge (>30 days), what form of accommodation can you accept?	□Sleeping on the floor □Bed-only □Tent □Partitioned cubicle □Don’t care □Other (please add)_______
19. When taking refuge in a basement for a long time (>30 days), what functions do you think the basement should have? (Multiple choices)	□Canteen □Game room □Chess and card room □Calligraphy and painting room □Gym □Coffee bar □Library □Open activity area □Psychological counseling □Health clinic □Supermarket
20. When taking refuge in a basement for a long time (>30 days), what facilities do you think can best alleviate the claustrophobic feeling in the basement?	□Skylight □Open courtyard (or patio) □Entrance sunken square □Central hall □High space
21. When taking refuge in a basement for a long time (>30 days) and cannot leave the basement for a long time, how do you most want to get information from the outside world?	□Go out and get it yourself □Get it through the Internet □Convey it through staff □Learn it through TV

. A total of 700 questionnaires were distributed, 565 collected, and 443 valid responses retained after eliminating errors and abnormal answering times (90 s–1200 s).

The valid respondents consisted of 186 males (42.0%) and 257 females (58.0%), with age distribution across under 18 (18.7%), 18–28 (36.8%), 29–50 (27.3%), and above 50 (17.2%). This basic profile confirms the sample aligns with the target population (urban residents potentially using underground refuges).

2.3.2. VR Experiment

This research employed a combination of mixed experimental design and orthogonal design methods commonly used in psychological research. Based on an orthogonal matrix, scenario variables and their treatment levels were systematically combined. Virtual environments were created using professional software and integrated into a VR experimental platform. A total of 130 volunteers participated, including 60 males and 70 females, aged 18–58 years (mostly 18–26 years), with bachelor’s (61.5%) or postgraduate (37.7%) education. They had varying VR experience: 10 with ≥3 experiences, 48 with 1–2, and 72 with none. Volunteers wore VR headsets to explore simulated scenarios, then completed a Likert scale questionnaire. Collected data were analyzed for results.

The eleven psychological indicators (Y1–Y11) in the VR experiment were designed as dependent variables to quantify user psychological responses, with foundations aligned to environmental psychology frameworks [37]. Selection followed two criteria: relevance to underground refuge contexts and consistency with validated assessment models.

Affective and interpersonal indicators: Y1 (Comfort), Y2 (Relaxation), Y3 (Affinity), and Y5 (Enjoyment) adapt established scales from social dynamics [38,39], stress reduction [40], and hedonic experience studies [41], with wording adjusted for underground-specific factors such as crowding and enclosure.

Spatial perception and functionality indicators: Y4 (Brightness Perception) and Y6–Y11 draw on underground environment research [42,43,44]. Y9 (Memorability) and Y10 (Sense of Order) are rooted in spatial legibility theory [44], while Y11 (Perceived Functionality) reflects usability engineering principles [45], addressing core psychological needs in enclosed emergency spaces.

Detailed mechanisms linking these indicators to spatial attributes are provided in Section 4.4.

3. Human Behavior, Psychological Responses, and Needs in the DUSR

3.1. Content and Design of the Questionnaire

People who are in the same situation due to accidental factors such as disasters are different from organized groups that are interdependent under normal circumstances. They have behavioral and psychological instability and increased aggressiveness [46]. Considering that underground space is suitable for emergency shelter (1 day) and medium- and long-term shelter (30–100 days), when researching underground space shelter, it is necessary to consider not only the behavior and psychological responses when a disaster occurs, but also the needs of long-term shelter in underground space: (1) In Wang et al.’s research, the refuge behaviors of affected people were classified into running toward open and well-lit routes, choosing familiar places for refuge, or displaying no refuge behavior [47]. The underground parking garage, as a familiar environment for nearby residents, can be investigated to understand their behavioral characteristics, habits, and perceptions related to refuge in underground spaces. (2) The refuge psychology of affected people was divided into instinctive reactions and fear-related psychological responses [48], and investigating their impressions of living in underground refuge spaces and their specific concerns can help identify their psychological fears in these spaces. (3)The typical refuge needs were classified into safety needs, physiological needs, and esteem needs [26,49]. Medium- and long-term refuge design requires an understanding of people’s preferences, needs, and psychological compensation factors related to accommodations and activity spaces in underground refuge areas.

The questionnaire was structured into four dimensions to investigate psychological factors related to underground refuge behavior, with design logic and theoretical basis detailed as follows (specific items and corresponding questions are listed in Table 2 to avoid redundancy):

Basic Information: Included demographic variables (age, gender, education, occupation) following standard survey practices, aiming to establish a basis for subgroup analysis (e.g., differences in needs across age groups). This dimension aligns with Table 2’s focus on capturing respondents’ fundamental characteristics (Question 1–5).

Behavioral Patterns: Informed by emergency response models [50,51], this dimension targeted refuge-related behaviors (e.g., disaster experience, shelter selection principles, willingness to enter underground spaces) to reflect real-world shelter-seeking tendencies. As outlined in Table 2, it corresponds to Question 6–12, with items refined to match disaster psychology frameworks [46].

Psychological Impressions: Drew from studies on emotional responses to underground environments [41], focusing on subjective perceptions (e.g., impressions of underground spaces, concerns about long-term refuge) to capture cognitive attitudes. Table 2 specifies its correspondence to Question 13–15, with options (e.g., “Enclosed/Safe”) validated by environmental psychology experts.

Environmental Needs: Based on Maslow’s hierarchy [52] and universal design principles [53], this dimension further integrated Altman’s privacy-regulation theory (1975) [54], which conceptualizes privacy as a dynamic balance between social interaction and seclusion. It explored accommodation preferences (e.g., ‘partitioned cubicles’ and ‘bed curtains’ in Q19), functional requirements, and psychological compensation factors for long-term refuge. As shown in Table 2, it includes Question 17–21, with items designed to reflect physiological and psychological needs [49].

All questions were refined based on rigorous alignment with the disaster psychology literature [46,47,48,49,50,51,52,53] to ensure clarity—for example, Q13 was revised from open-ended to specific options (e.g., ‘Enclosed/Safe’) by referencing validated scales in environmental behavior studies [41], avoiding ambiguity.

Regarding reliability, traditional metrics (e.g., Cronbach’s α) are inapplicable for yes/no/multiple-choice questions. Instead, data reliability was ensured through the following: (1) alignment with disaster response models [50,51], Maslow’s hierarchy [52], and environmental psychology [41]; (2) exclusion of 122 invalid responses via Q16 screening and <90 s completion times (based on 20-participant pre-test), retaining 443 samples; (3) logical consistency between behavioral (Q9–Q11) and psychological (Q13) items (e.g., refuge willingness correlated with ‘Safe’ perceptions).

Table 2. Questionnaire dimensions, purpose and corresponding items.

No.	Questionnaire Dimensions	Setting Purpose	Corresponding Questionnaire Items
(1)	Basic Information	Understand the basic information of the survey subjects	(Question 1–5) Age, gender, education level, occupation, whether there is experience of living or working underground
(2)	Behavioral Patterns	The respondents’ experience, habits, cognition and other behavioral characteristics of underground space refuge	(Question 6–12) Whether they have experienced a particular disaster, the principles for choosing a refuge site, whether they know the location of the city’s emergency refuge sites, whether they are willing and under what conditions they are willing to enter the underground space for refuge, and the activity choices for medium-term and long-term refuge
(3)	Psychological Impressions	The respondents’ impressions of life in underground refuge spaces and the specific content of their concerns	(Question 13–15) Impressions of underground spaces and concerns about life in ground and underground spaces
(4)	Environmental Needs	The respondents’ willingness, needs and psychological compensation factors for the accommodation and activity space of underground refuge spaces	(Question 17–21) Accommodation layouts, activity choices, functional requirements, psychological compensation facilities and ways to obtain external information during medium-term and long-term and long-term refuge

Table 2. Questionnaire dimensions, purpose and corresponding items.

No.	Questionnaire Dimensions	Setting Purpose	Corresponding Questionnaire Items
(1)	Basic Information	Understand the basic information of the survey subjects	(Question 1–5) Age, gender, education level, occupation, whether there is experience of living or working underground
(2)	Behavioral Patterns	The respondents’ experience, habits, cognition and other behavioral characteristics of underground space refuge	(Question 6–12) Whether they have experienced a particular disaster, the principles for choosing a refuge site, whether they know the location of the city’s emergency refuge sites, whether they are willing and under what conditions they are willing to enter the underground space for refuge, and the activity choices for medium-term and long-term refuge
(3)	Psychological Impressions	The respondents’ impressions of life in underground refuge spaces and the specific content of their concerns	(Question 13–15) Impressions of underground spaces and concerns about life in ground and underground spaces
(4)	Environmental Needs	The respondents’ willingness, needs and psychological compensation factors for the accommodation and activity space of underground refuge spaces	(Question 17–21) Accommodation layouts, activity choices, functional requirements, psychological compensation facilities and ways to obtain external information during medium-term and long-term and long-term refuge

3.2. Questionnaire Results and Discussion

Table 3. Basic information of the respondents.

Item	Option	Number	Proportion
Gender	Male	186	42.0%
	Female	257	58.0%
Age	Under 18	83	18.7%
	18–28	163	36.8%
	29–50	121	27.3%
	Above 50	76	17.2%
Education level	Junior high School and below	74	16.7%
	High school and Junior college	111	25.1%
	Undergrad	175	39.5%
	Undergrad above	83	18.7%
Profession	Civil servants	12	2.7%
	Staff	98	22.1%
	Worker	35	7.9%
	Merchant	17	3.8%
	Student	168	37.9%
	Teacher	25	5.6%
	Unemployed	21	4.7%
	Other	67	15.1%
Refugee experience	Yes	124	28.0%
	No	319	72.0%
Selection criteria	Nearby	153	34.5%
	Safety	188	42.5%
	Follow the crowd	44	9.9%
	Good environment	57	12.8%
	Other	1	0.3%
Do you know where the shelter is?	Yes	114	25.7%
	No	329	74.3%
Are you willing to enter the underground shelter?	Yes	366	82.6%
	No	77	17.4%

details the full demographic characteristics of the 443 valid respondents, covering diverse groups across multiple dimensions. This distribution aligns with the target population (urban residents potentially using underground refuges), supporting the generalizability of findings.

After analyzing the questionnaire data, results were obtained. Here, Question 14 “During the medium- to long-term (about 30 days or more) ground refuge life, what aspects will you worry about?” and Question 15 “If you stay in the underground space for a medium to long term (about 30 days or more), what aspects will you worry about?” are used as examples to demonstrate the analytical process, which was carried out using SPSS software.

As shown in Figure 2, a cross-tabulation chi-square test was conducted to compare people’s concerns regarding surface-level and underground refuge living, revealing no significant differences between the two. Among the concerns, “Drink water” and “Dine” ranked slightly higher, followed by “power supply”, “Internet”, “Toilet facilities” and “Informed outside”. Approximately 9% of respondents expressed worries about “Wash”, while only about 4.0% were concerned about “Amusement.” A few individuals additionally noted concerns regarding “sanitation conditions” and “air quality.” This indicates that physiological needs are people’s primary concerns, encompassing not only water and food but also spatial functions, environmental sanitation, and air quality. Secondary concerns involve information acquisition channels, while entertainment and activities ranked last.

Furthermore, chi-square tests on cross-tabulations between concerns about surface and underground refuge living and basic demographic variables showed p-values greater than 0.05, indicating no significant associations with age, gender, education level, occupation, or experience living and working in underground spaces.

Following the above methodology, the overall results and analysis of the questionnaire survey are summarized as follows.

At the level of behaviors, most people lack disaster prevention knowledge, and only a few know the locations of emergency shelters announced by the city. When a disaster strikes, most people are willing to take refuge in underground spaces, among which young people and people with higher education levels are more willing, especially when “terrorist attacks”, “meteorological disasters”, and “nuclear accidents” occur. If clear signs are set up, most people will take refuge in underground spaces. The main principles for choosing a shelter are safety and proximity, followed by a good environment and herd mentality. Regarding activity needs, most people tend to participate in public activities; some choose to lie down to rest or take a walk, and a few prefer activities such as “playing video games” or “reading”.

In terms of psychological responses, most people’s impression of underground space is “closed”, “dark”, and “depressing”, followed by “fear”, “unsafe”, and “strange”. Among them, women and people with higher education have a more negative impression of underground space. A few people think underground space is “safe”, which may be based on its physical environment characteristics. In terms of worry, people have fear and anxiety about refuge life. They are most concerned about “drinking water” and “eating”, followed by “power supply”, “mobile phone signal and network”, “going to the toilet”, and “learning about the outside world”. Some people are also worried about “washing” or “entertainment”, and very few people add “sanitary conditions” and “air quality” issues.

Regarding needs, most people prefer “personal private space” in accommodation space. People aged 18–28, those with higher education, and those who are civil servants, teachers, students, and company employees are more inclined to choose private space. In terms of accommodation, people aged 18–28 and 29–50 prefer “partitioned cubicle”, people under 18 prefer “tents”, and people over 50 prefer “bed-only” or “it doesn’t matter”. Women and those with higher education prefer “partitioned cubicle”, while those with lower education prefer “bed-only”. In terms of functional requirements, the most popular requirements are to add “health center”, “canteen”, “psychological consultation”, “open activity area”, and other functions. A few people suggested adding a “chess and card room”, “calligraphy and painting room”, and other facilities. For psychological compensation measures, a semi-open atrium with reinforced transparent roofing was identified as the most effective spatial intervention to alleviate claustrophobic feelings. This approach preserves the visual openness and natural light exposure typically associated with open courtyards, while ensuring disaster-resilient structural integrity through the use of explosion-proof and shatter-resistant materials, in compliance with GB51143-2015 shelter design standards. Additional measures—such as full-height interiors, skylights, and sunken entrance squares—can also enhance psychological comfort when similarly implemented with protective reinforcements.

Questionnaire findings guided Section 4 design by linking three key insights to VR variables: behavioral data (Q7, Q9–Q11) highlighted ‘safety’ and ‘guidance’ as critical drivers, directly informing the design of X5 (accommodation) and X9 (signage); psychological impressions (Q13–Q15) shaped Y variables by establishing correlations between fear responses and spatial features like X9; environmental needs (Q19–Q20) validated X10 (greenery) and X11 (skylight) as essential for enhancing functionality and openness in refuge spaces.

4. Key Psychological Impact Factors in the DUSR

4.1. Experimental Procedure

This research uses the semantic analysis method and VR experimental method commonly used in psychology to obtain people’s psychological perception and subjective evaluation of space, and explore the main influencing factors and psychological index factors of underground space on people. First, this research obtains various physical indicators of existing underground space through field research and literature review, determines the range of spatial indicators that affect psychology through analysis, and proposes the values of various physical space variables (independent variables). Second, orthogonal experimental design (Section 4.3) was used to generate 16 scenario combinations, which were imported into the VR platform after modeling. Finally, data analysis was conducted using SPSS 23.0. Third, this research determined the dependent variables of psychological indicators through literature review and preliminary experiment analysis, established a Likert scale, organized volunteers to conduct VR experiments, and collected data through scale questionnaires. Finally, data analysis was conducted using SPSS 23.0, and the results were used to derive the research’s conclusions. The experimental module design is shown in Figure 3.

4.2. Selection Rationale and Level Definition of Independent Variables

The 11 physical environmental variables (X1–X11) selected in this study focus on visual–spatial attributes, covering three core dimensions: physical space parameters (X1–X4), functional layout design (X5–X6), and perceptual environment factors (X7–X11). This classification aligns with environmental psychology frameworks [37], which emphasize the interplay of spatial form, functional organization, and sensory perception in shaping user responses. All variables are both theoretically grounded and technically manipulable in VR environments.

(1)

X1—Effective refuge area: Guided by the Code (GB51143-2015) [18], which stipulates dormitory group areas < 1080 m² and clusters < 4320 m², combined with fire protection (3000–4000 m²) [55] and civil defense (700–2000 m²) [9] standards. Four levels are set: 1000 m², 2000 m², 3000 m², and 4000 m².

(2)

X2—Clear ceiling height: Selected for its impact on psychological comfort and perceived spaciousness in underground environments [41,56]. Based on typical ranges (residential underground: 2–2.6 m; public underground: 2–5.9 m), two levels are used: 2.3 m and 3.8 m.

(3)

X3—Public corridor width: Supported by research on circulation efficiency and psychological ease in underground complexes [41]. Referencing the Code [36] (minimum sidewalk width 1.5 m) and general office standards (2.4 m), two levels are compared: 1.5 m and 2.4 m.

(4)

X4—Per capita effective refuge area: Derived from the Code [18] (≥3.0 m²/person for medium- to long-term use), with increments of 1.5 m²/person: 3.0 m²/person, 4.5 m²/person, 6.0 m²/person, and 7.5 m²/person.

(5)

X5—Accommodation layout: Informed by Section 3 survey results, two levels are compared: dormitory units with beds vs. units with standard disaster relief tents (3.7 × 3.2 m, eaves height 1.75 m, ridge height 2.30 m).

(6)

X6—Spatial geometry: Linked to visual cognition and spatial stress [41,57]. Based on survey data, two levels are set: strictly functional division of beds/tents vs. integrated “channels, nodes, areas, boundaries, and landmarks” (public–private combination).

(7)

X7—Interior color and X8 Indoor illuminance: Both selected for their influence on emotional responses in enclosed spaces [41,56]. For X7, color-coded functional blocks vs. a single uniform color. For X8, referencing standards [57] (garage: 50 lx; dormitory/living room: 100 lx), two levels are used: 50 lx and 100 lx.

(8)

X9—Signage system: Derived from Section 3 questionnaire findings, reflecting user needs for orientation. Two levels are compared: presence vs. absence of signage (functional divisions, directional indicators, location markers).

(9)

X10—Greenery layout: Supported by environmental psychology [38,41] and VR experiments on stress mitigation. Two levels are compared: centralized “garden” areas vs. scattered potted plants.

(10)

X11—Skylight: Selected for its impact on brightness perception and psychological relief [41]. Two levels are compared: presence vs. absence of skylights (4 m², referencing underground garage lighting standards).

4.3. Orthogonal Experimental Design

The ideal scene model was a square plane (8.4 m × 8.4 m column span) with X1-based grouping. X2 clear ceiling height (2.3 m/3.8 m) was determined by controlling total height (beams + pipelines) at 1.3 m to ensure consistency with real-world underground structural constraints, corresponding to floor heights of 3.6 m and 5.1 m (Figure 4).

To manage 11 independent variables (2 with 4 levels, 9 with 2 levels), orthogonal design was used to reduce experiments. Via SPSS, an L4²2⁹ orthogonal table generated 16 scenario combinations (

Table 4. Orthogonal experiment independent variable combination table L4²2⁹.

Scene	Group	X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	X11
S1	C	3000 m2	2.3 m	2.4 m	7.5 m2	Bed	No	Yes	100 lx	Yes	Dispersed	No
S2	B	2000 m2	2.3 m	1.5 m	7.5 m2	Tent	No	No	50 lx	Yes	Centralized	Yes
S3	B	2000 m2	3.8 m	2.4 m	6.0 m2	Bed	No	No	100 lx	No	Dispersed	Yes
S4	A	1000 m2	3.8 m	2.4 m	4.5 m2	Tent	No	No	100 lx	Yes	Centralized	No
S5	B	2000 m2	2.3 m	2.4 m	4.5 m2	Bed	Yes	Yes	50 lx	No	Centralized	No
S6	D	4000 m2	2.3 m	2.4 m	3.0 m2	Tent	No	Yes	100 lx	No	Centralized	Yes
S7	D	4000 m2	2.3 m	1.5 m	6.0 m2	Bed	Yes	No	100 lx	Yes	Centralized	No
S8	C	3000 m2	3.8 m	1.5 m	6.0 m2	Tent	No	Yes	50 lx	No	Centralized	No
S9	D	4000 m2	3.8 m	2.4 m	7.5 m2	Tent	Yes	No	50 lx	No	Dispersed	No
S10	A	1000 m2	3.8 m	1.5 m	7.5 m2	Bed	Yes	Yes	100 lx	No	Centralized	Yes
S11	A	1000 m2	2.3 m	2.4 m	6.0 m2	Tent	Yes	Yes	50 lx	Yes	Dispersed	Yes
S12	D	4000 m2	3.8 m	1.5 m	4.5 m2	Bed	No	Yes	50 lx	Yes	Dispersed	Yes
S13	C	3000 m2	3.8 m	2.4 m	3.0 m2	Bed	Yes	No	50 lx	Yes	Centralized	Yes
S14	A	1000 m2	2.3 m	1.5 m	3.0 m2	Bed	No	No	50 lx	No	Dispersed	No
S15	B	2000 m2	3.8 m	1.5 m	3.0 m2	Tent	Yes	Yes	100 lx	Yes	Dispersed	No
S16	C	3000 m2	2.3 m	1.5 m	4.5 m2	Tent	Yes	No	100 lx	No	Dispersed	Yes

), ensuring efficient factor analysis. Each scene of experiments is divided into four groups: A, B, C, and D, based on the building area. At the same time, each scene was drawn with a plan, as shown in Figure 5. After controlling some independent variable indicators, each scene was modeled through Sketchup and then imported into Twinmotion 2022.1 for material assignment and rendering to generate a VR scene, as shown in Figure 6. Finally, the VR device was connected through SteamVR, and the experiment was conducted using Twinmotion Presenter.

To clarify the VR simulation technical set up and its alignment with variable measurement, key parameters are specified as follows: The experiment adopted the HTC Vive system with a headset resolution of 2160 × 1200 and a 90 Hz refresh rate, connected to the host via a wired connection for stable performance. A wired dual-mode precise positioning tracking system supported room-scale movement (2 m × 2 m tracking area) with controller-based navigation (joystick for locomotion, motion controllers for orientation), enabling natural free movement and accurate spatial localization within the simulated scene. The visual perspective was set to a first-person view consistent with daily life (horizontal field of view 90°, eye height 1.65 m), ensuring natural perception of spatial attributes (e.g., X2). Simulation content included typical underground refuge layouts with predefined physical parameters (e.g., X1, X5, X8). After scene exploration, participants completed the psychological evaluation scale (Table 5) via physical paper questionnaires to collect Y1–Y11 data.

4.4. Selection of Dependent Variables

The eleven dependent variables (Y1–Y11) are developed through two approaches: adapting existing validated scales and creating new items based on theoretical frameworks, forming a psychological evaluation framework. Specifically, Y1 (Comfort), Y2 (Relaxation), Y3 (Affinity), and Y5 (Enjoyment) draw on established scales measuring social dynamics [38,39], stress reduction [40], and hedonic experience [41], with wording adjusted for underground refuge contexts. The remaining indicators (Y4, Y6–Y11) are newly created by synthesizing empirical studies on underground environments [42,43,44,45], tailored to spatial and psychological cues unique to refuge settings.

“Y1 Comfort” and “Y3 Affinity”, which reflect emotional and interpersonal responses in subterranean settings, are supported by research on social dynamics in underground workspaces [38,39,58]. Drawing conceptual parallels with “vertical communities” [58], emerging discussions on “underground communities” highlight similar challenges in fostering safety and psychological well-being under conditions of spatial enclosure, artificial lighting, and limited social visibility. These socio-environmental dynamics are central to comfort perception in confined underground spaces. “Y2 Relaxation” is grounded in studies highlighting the role of built environments in stress reduction and mental ease, with elements such as natural light shown to facilitate psychological relaxation [40]. “Y4 Brightness Perception” synthesizes insights from Chinese research on environmental impacts in underground spaces [42] and international studies on lighting’s effect on well-being [55], underscoring its importance in visual comfort and mood regulation.

“Y5 Enjoyment” captures the hedonic aspects of spatial experience, as emphasized in reviews of environmental design factors influencing psychological health in subterranean environments [39,41]. “Y6 Spatial Richness” and “Y7 Spatial Openness” relate to perceptions of environmental complexity and spaciousness, drawing on foundational environmental psychology [44], user-centered design evaluations of underground malls [59], and VR-based investigations into stress responses to architectural configurations [60]. “Y8 Height Perception” is informed by studies that measure visual cues for ceiling height and the influence of surface brightness and lighting [61,62,63].

“Y9 Memorability” and “Y10 Sense of Order”, rooted in spatial legibility theory [44], reflect cognitive mapping and environmental coherence. “Y11 Perceived Functionality”, following the principles of usability engineering [45], evaluates users’ judgments of spatial effectiveness and functional adequacy.

Together, these variables establish a multidimensional framework for assessing psychological impressions of underground refuge environments via immersive VR simulation, bridging theoretical foundations with empirical validity, and formulates a psychological evaluation indicator scale as shown in Table 5.

Table 5. Psychological evaluation index scale.

Indicator	Negative	Quantitative Degree							Positive
Y1 Comfort	Uncomfortable	1	2	3	4	5	6	7	Comfortable
Y2 Relaxation	Depressed	1	2	3	4	5	6	7	Relaxed
Y3 Affinity	Alienation	1	2	3	4	5	6	7	Belonging
Y4 Brightness Perception	Dark	1	2	3	4	5	6	7	Bright
Y5 Enjoyment	Boring	1	2	3	4	5	6	7	Interesting
Y6 Spatial Richness	Monotonous	1	2	3	4	5	6	7	Rich
Y7 Spatial Openness	Crowded	1	2	3	4	5	6	7	Open
Y8 Height Perception	Low	1	2	3	4	5	6	7	Tall
Y9 Memorability	Difficult	1	2	3	4	5	6	7	Easy
Y10 Sense of Order	Chaotic	1	2	3	4	5	6	7	Orderly
Y11 Functional completeness	Functionally lacking	1	2	3	4	5	6	7	Functionally Complete

Table 5. Psychological evaluation index scale.

Indicator	Negative	Quantitative Degree							Positive
Y1 Comfort	Uncomfortable	1	2	3	4	5	6	7	Comfortable
Y2 Relaxation	Depressed	1	2	3	4	5	6	7	Relaxed
Y3 Affinity	Alienation	1	2	3	4	5	6	7	Belonging
Y4 Brightness Perception	Dark	1	2	3	4	5	6	7	Bright
Y5 Enjoyment	Boring	1	2	3	4	5	6	7	Interesting
Y6 Spatial Richness	Monotonous	1	2	3	4	5	6	7	Rich
Y7 Spatial Openness	Crowded	1	2	3	4	5	6	7	Open
Y8 Height Perception	Low	1	2	3	4	5	6	7	Tall
Y9 Memorability	Difficult	1	2	3	4	5	6	7	Easy
Y10 Sense of Order	Chaotic	1	2	3	4	5	6	7	Orderly
Y11 Functional completeness	Functionally lacking	1	2	3	4	5	6	7	Functionally Complete

4.5. Experimental Process

In order to meet the requirement of no less than 30 subjects in each group of psychology experiments, this research recruited a total of 130 subjects with good vision and mental state, divided into Group A (32 subjects), Group B (33 subjects), Group C (32 subjects), and Group D (33 subjects). The subjects comprised 60 males and 70 females. Educational attainment was predominantly at the bachelor’s (61.5%) and postgraduate (37.7%) levels. Ages ranged from 18 to 58 years, with overrepresentation in the 18–26 demographic, shown in

Table 6. Age frequency table of subjects.

Age	18	19	20	21	22	23	24	25	26	27	28	30	48	58
Number	18	19	12	11	15	15	16	11	5	2	1	3	1	1

. The subjects had 10 people who had 3 or more VR experiences, 48 people who had 1–2 experiences, and 72 people who had no VR experience.

Before the experiment, the experimenter recorded the basic information of the subjects, introduced the experimental process and purpose, guided the operation of the handle, and let them adapt to the virtual scene. In order to avoid too many subjects, the experiment adopted a mixed design, and each subject participated in only one group of experiments. The experimental process is shown in Figure 7. The total duration was 30 min, including four rounds of scene exploration—one selected scenario per group—followed by the completion of a Likert scale questionnaire after each exploration.

4.6. Results and Analysis

To analyze the psychological experimental data, this research adopted several widely used quantitative methods, including mean analysis, correlation analysis, and multiple linear regression [64,65,66]. These methods were applied to identify distribution patterns, examine inter-variable associations, explore predictive factors, and extract latent dimensions within the psychological variables.

4.6.1. Experimental Scene Indicator Mean Evaluation

Due to the large sample size (520), this research used P-P graphs and histograms to conduct a preliminary analysis of the participants’ psychological index data and found that the data in each group were close to a normal distribution. At the same time, the reliability of the data was analyzed, and Cronbach’s alpha of the scale was found to be 0.887 (>0.8), which has good credibility and can be further analyzed without revision. To ensure the robustness of our findings, we conducted covariate analyses to examine potential influences of demographic, experiential, and vulnerability-related factors on psychological outcomes. Key covariates included age, gender, education level, academic major, frequency of VR experience, and indicators of potential vulnerabilities (e.g., self-reported mobility limitations or chronic health conditions). Statistical tests (ANOVA for age, major, VR experience, and vulnerability indicators; independent samples t-test for education level and gender) revealed no significant associations between these covariates and the primary psychological indicators measured (all p > 0.05). Specifically, (1) age did not exhibit a statistically significant effect on perceptions of comfort, safety, or usability; (2) vulnerability-related factors (e.g., mobility limitations) also showed no significant correlation with psychological responses. These results suggest that the core relationships identified between spatial design factors (e.g., ceiling height, per capita area) and psychological outcomes are consistent across different age groups and populations with varying vulnerability levels within our sample framework.

The mean of the psychological index data of each scene was calculated and organized into a data mean table (

Table 7. Mean values of dependent variables for each scene.

Scene	Y1	Y2	Y3	Y4	Y5	Y6	Y7	Y8	Y9	Y10	Y11
S1	4.78	5.03	4.91	6.06	4.81	4.47	3.66	3.72	5.31	5.03	4.59
S2	4.52	4.42	4.58	5.18	3.91	3.88	3.58	2.97	4.82	5.09	3.82
S3	4.39	4.55	4.97	5.97	3.82	3.52	3.67	5.21	5.42	4.94	3.55
S4	4.59	4.38	4.50	5.63	4.28	4.16	3.63	4.84	4.88	4.94	4.22
S5	4.00	3.91	4.64	5.00	4.09	4.42	3.18	3.27	5.15	4.18	4.12
S6	3.00	2.85	3.52	4.42	2.85	2.67	1.97	2.94	2.42	3.91	3.06
S7	4.45	4.21	4.64	5.27	4.12	4.21	3.61	3.36	5.09	5.00	4.39
S8	4.47	4.06	3.84	5.09	4.44	4.41	3.38	4.59	4.34	4.94	4.31
S9	5.30	4.97	4.73	5.18	4.73	4.97	4.27	4.61	4.42	5.36	5.33
S10	4.91	4.78	4.88	5.75	4.16	4.13	4.38	4.84	6.13	5.38	4.44
S11	5.22	4.59	4.88	4.69	4.72	4.88	3.88	3.44	5.03	5.31	5.06
S12	4.24	4.30	4.27	5.39	3.79	3.97	3.48	4.55	5.36	4.91	4.24
S13	3.81	3.59	4.34	4.03	4.00	3.91	2.91	4.03	4.84	4.44	3.59
S14	3.00	3.00	3.78	4.13	2.88	2.47	1.59	3.06	4.38	3.88	2.66
S15	3.76	3.64	4.55	5.21	3.64	3.88	2.03	4.33	3.61	4.39	3.61
S16	4.03	3.72	4.22	4.63	4.00	3.84	2.84	3.38	2.69	3.91	4.16

). Dark gray cells are data with a mean above 4.5, light gray cells with a score below 3.5, and white cells are data between 3.5 and 4.5. From a horizontal analysis, scenes S1, S2, S3, S9, S10, and S11 have more indicators with a score higher than 4.5, indicating that factors such as higher clear ceiling height, more effective refuge area, more expansive public passages, spatial imagery, interior color distinction, and scattered plant arrangement may bring more positive experiences; in comparison, scenes S6 and S14 have more indicators with a score lower than 3.5, which may be related to factors such as lower clear ceiling height and smaller per capita effective area. At the same time, the longitudinal analysis shows that psychological indicators Y1, Y3, Y4, Y9, and Y10 appear more frequently in high-scoring scenes, indicating that these dependent variables may be closely related to the above-mentioned independent variables. Y7 appears most frequently in low-scoring scenes, which may be related to independent variables X3 and X4.

According to the mean table, this research found that the scenes with the highest scale evaluation were S9, S11, S10, and S1, and the lowest were S14 and S6. Factors such as greater ceiling height, larger per capita effective refuge area, wider public corridor width, obvious spatial geometry, differentiated interior color schemes, and dispersed greenery layout were associated with more positive user perceptions. From the perspective of a single psychological indicator, most scenes are considered crowded but orderly, and all scenes are generally brighter, which may be related to the high level of processing of the independent variables. Overall, there is a specific correlation between multiple independent variables of underground refuge space and the dependent variable of psychological indicators.

4.6.2. Correlation Analysis

This research adopted an orthogonal experimental design and only studied the correlation between the independent and dependent variables and between the dependent variables. The strength of the correlation was determined by the Pearson correlation coefficient (r) and significance (p). r between 0.100 and 0.200 was considered a low correlation, between 0.200 and 0.400 was considered a moderate correlation, and above 0.400 was considered a high correlation. When p < 0.05, it was considered to be significantly correlated, and when p < 0.01, it was considered highly significantly correlated. The analysis results of the independent and dependent variables are shown in

Table 8. Results of the relevance analysis between independent variables and dependent variables.

		Y1	Y2	Y3	Y4	Y5	Y6	Y7	Y8	Y9	Y10	Y11
X1	r	−0.037	−0.027	−0.096 *	−0.029	0.002	0.029	−0.001	−0.044	−0.181 **	−0.026	0.072
	p	0.401	0.532	0.029	0.514	0.957	0.507	0.990	0.314	0.000	0.557	0.099
X2	r	0.121 **	0.116 **	0.050	0.141 **	0.066	0.092 *	0.137 **	0.476 **	0.149 **	0.144 **	0.065
	p	0.006	0.008	0.254	0.001	0.134	0.036	0.002	0.000	0.001	0.001	0.137
X3	r	0.082	0.077	0.089 *	0.015	0.104 *	0.095 *	0.090 *	0.043	0.037	0.028	0.085
	p	0.061	0.081	0.043	0.733	0.018	0.031	0.039	0.327	0.399	0.520	0.053
X4	r	0.423 **	0.395 **	0.221 **	0.292 **	0.272 **	0.278 **	0.423 **	0.114 **	0.296 **	0.324 **	0.330 **
	p	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.009	0.000	0.000	0.000
X5	r	0.061	−0.035	−0.086	−0.078	0.038	0.068	−0.038	−0.043	−0.347 **	0.004	0.088 *
	p	0.163	0.428	0.051	0.076	0.383	0.123	0.383	0.327	0.000	0.919	0.046
X6	r	−0.121 **	−0.038	−0.131 **	0.102 *	−0.120 **	−0.208 **	−0.085	0.027	−0.001	−0.016	−0.193 **
	p	0.006	0.392	0.003	0.020	0.006	0.000	0.052	0.540	0.980	0.709	0.000
X7	r	−0.007	−0.010	0.018	−0.075	−0.030	−0.078	0.011	−0.008	−0.026	−0.019	−0.074
	p	0.865	0.824	0.686	0.088	0.493	0.074	0.800	0.854	0.557	0.660	0.093
X8	r	−0.034	0.010	0.057	0.204 **	−0.041	−0.092 *	−0.024	0.091 *	−0.104 *	−0.031	−0.054
	p	0.434	0.824	0.197	0.000	0.350	0.036	0.593	0.037	0.017	0.477	0.217
X9	r	−0.109 *	−0.105 *	−0.108 *	−0.063	−0.101 *	−0.127 **	−0.058	0.030	−0.145 **	−0.127 **	−0.085
	p	0.013	0.017	0.013	0.152	0.021	0.004	0.185	0.501	0.001	0.004	0.053
X10	r	−0.049	−0.074	−0.073	−0.045	−0.025	−0.011	0.046	−0.070	0.051	0.004	−0.057
	p	0.261	0.092	0.097	0.306	0.575	0.805	0.297	0.111	0.250	0.919	0.194
X11	r	0.013	0.018	−0.002	0.072	0.079	0.100 *	−0.053	0.019	0.017	−0.007	0.063
	p	0.760	0.680	0.971	0.101	0.070	0.022	0.226	0.668	0.702	0.866	0.154

* The correlation is significant at the 0.05 level. ** The correlation is significant at the 0.01 level.

. The light gray cells represent r absolute values less than 0.100, and the dark gray cells represent r absolute values greater than 0.100. Among the underground space indicators, the three independent variables of interior color, greenery layout, and skylight have no significant correlation with psychological indicators. The four related psychological indicators are clear ceiling height, per capita effective refuge area, signage system, and spatial geometry. Among them, per capita effective refuge area is the most critical, with a low or above positive correlation with 11 psychological indicators, and a medium or above positive correlation with 10 psychological indicators.

The results of the dependent variable and dependent variable analysis are shown in

Table 9. Correlation between dependent variables.

		Y1	Y2	Y3	Y4	Y5	Y6	Y7	Y8	Y9	Y10	Y11
Y1	r	1
	p
Y2	r	0.753 **	1
	p	0.000
Y3	r	0.483 **	0.515 **	1
	p	0.000	0.000
Y4	r	0.443 **	0.511 **	0.419 **	1
	p	0.000	0.000	0.000
Y5	r	0.617 **	0.544 **	0.473 **	0.390 **	1
	p	0.000	0.000	0.000	0.000
Y6	r	0.604 **	0.524 **	0.434 **	0.367 **	0.726 **	1
	p	0.000	0.000	0.000	0.000	0.000
Y7	r	0.635 **	0.638 **	0.370 **	0.353 **	0.522 **	0.500 **	1
	p	0.000	0.000	0.000	0.000	0.000	0.000
Y8	r	0.321 **	0.383 **	0.244 **	0.406 **	0.284 **	0.248 **	0.410 **	1
	p	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Y9	r	0.341 **	0.373 **	0.283 **	0.289 **	0.255 **	0.216 **	0.342 **	0.282 **	1
	p	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Y10	r	0.473 **	0.445 **	0.335 **	0.354 **	0.336 **	0.314 **	0.404 **	0.295 **	0.513 **	1
	p	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Y11	r	0.575 **	0.527 **	0.400 **	0.352 **	0.564 **	0.661 **	0.445 **	0.235 **	0.227 **	0.411 **	1
	p	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000

** The correlation is significant at the 0.01 level.

. Light gray cells represent data with significant correlation, and dark gray cells represent data with significant correlation and correlation coefficients greater than 0.600. In addition, there is a significant positive correlation of 0.01 level between the dependent variable psychological indicators, and r is more significant than 0.200, reaching a medium or above correlation. Among them, the correlation coefficients of Y1 with Y2, Y5, Y6, and Y7 are all over 0.600, and there is a strong positive correlation. The correlation coefficient between Y2 and Y7 is above 0.600, with a strong positive correlation. The correlation coefficient between Y5 and Y6 is above 0.700, with a strong positive correlation. The correlation coefficient between Y6 and Y11 is above 0.600, with a strong positive correlation. The five psychological indicators of comfort, relaxation, enjoyment, spatial richness, and spatial openness are strongly correlated.

4.6.3. Multiple Linear Regression Analysis

This research estimated the curves of a single independent variable and related dependent variables and found that linear fitting was the best method for further exploring the quantitative relationship between independent and dependent variables. Therefore, multiple linear regression analysis was used for subsequent research. The corresponding relationship between the independent and dependent variables has been obtained through previous experiments, as shown in

Table 10. Correlation between independent variables and dependent variables.

Dependent Variables	Y1	Y2	Y3	Y4	Y5	Y6	Y7	Y8	Y9	Y10	Y11
Independent Variables	X2 X4 X6 X9	X2 X4 X9	X4 X6 X9	X2 X4 X6 X8	X3 X4 X6 X9	X4 X6 X9	X2 X4	X2 X4	X1 X2 X4 X5 X8 X9	X2 X4 X9	X4 X6

. The 11 dependent variables were subjected to multiple linear regression analysis, and after validity judgment, it was found that all met the requirements.

Taking Y1 as an example, we can obtain the linear regression Equation (1) through analysis:

$$\mathrm{Y}1=0.208\mathrm{X}2+0.325\mathrm{X}4+0.312\mathrm{X}6+0.281\mathrm{X}9+2.830$$

(1)

From the Equation, we can see that X4 has the most significant impact on Y1; for every 1 m² increase in X4, the Y1 index score increases by 0.325. The second is X6. When X6 takes “1” (obvious spatial geometry), the Y1 index increases by 0.312. Then comes X9. When X9 takes “1” (there are signage systems), the Y1 index increases by 0.281. This aligns with Appleyard’s (1970) spatial behavior theory [67], which emphasizes that ‘channels, nodes, and landmarks’ (X6) and clear signage (X9) enhance spatial legibility, reducing navigational stress and thus improving comfort. Finally, X2. For every 1 m increase in clear ceiling height, the Y1 index increases by 0.208. This alignment with X6 (spatial geometry) and X9 (signage system) echoes core principles of CPTED (Crime Prevention Through Environmental Design): obvious spatial geometry enhances ‘natural surveillance’ by reducing visual blind spots, while clear signage systems improve ‘legibility’—both contributing to reduced anxiety and enhanced comfort in enclosed underground spaces. Therefore, the trend of the dependent variable can be predicted based on the independent variable, and the range of values of the independent variable can also be predicted given the dependent variable. When Y1 is more significant than 4.0, according to the definition of the seven-level Likert scale, it can be considered that the psychological feeling in the underground space reaches “average” or above. The setting of spatial geometry and signage system can enhance the comfort of underground space. At the same time, when the minimum clear ceiling height is 2.0 m, the per capita effective refuge area must reach 4.1 m² or more to ensure that the comfort level is above 4.0. In addition, when the per capita adequate refuge area is at least 3.0 m², the clear ceiling height must reach 3.8 m or more to meet the comfort requirements.

In this research, comfort (Y1) in underground refuge spaces is not merely a numerical threshold but a multidimensional construct aligned with the specific demands of emergency shelter contexts. It encompasses three key dimensions derived from the regression analysis of influencing factors (X4, X6, X9): (1) Freedom from crowding: Relieved by sufficient per capita effective refuge area (X4), which reduces spatial constraint. (2) Spatial intelligibility: Enhanced by distinct spatial geometry (X6), which improves orientation in enclosed underground environments. (3) Functional reassurance: Supported by signage systems (X9) that provide navigational guidance, critical for reducing anxiety in high-stress scenarios.

The 4.0 cutoff on the 7-point Likert scale (where 1 = ‘extremely uncomfortable’ and 7 = ‘extremely comfortable’) thus represents a practical marker: scores above 4.0 indicate that participants perceived the space as meeting or exceeding basic needs across these dimensions, reflecting a psychologically acceptable state for refuge users.

The remaining 10 regression equations were sorted out and predicted, as shown in

Table 11. Multiple linear regression equations and prediction results.

Y2		Y2 = 0.213X2 + 0.325X4 + 0.288X9 + 1.627	(2)
	prediction	① Setting up a signage system can enhance the sense of ease in underground space; ② When the minimum clear ceiling height is 2.0 m, the per capita effective refuge area must be at least 5.1 m2 to ensure a sense of relaxation above 4.0; ③ When the minimum per capita effective refuge area is 3.0 m2, the clear ceiling height must be at least 5.3 m to ensure a sense of comfort above 4.0.
Y3		Y3 = 0.156X4 + 0.312X6 + 0.258X9 + 3.346	(3)
	prediction	① Setting up a spatial geometry and signage system can significantly enhance the affinity with the underground space; ② When the minimum per capita effective refuge area is 3.0 m2, at least one spatial geometry and signage system must be set to ensure that the affinity is above 4.0.
Y4		Y4 = 0.241X2 + 0.225X4 − 0.262X6 + 0.004X8 + 2.880	(4)
	prediction	① Not setting up spatial geometry can avoid reducing the brightness perception index value of underground space; ② The effect of indoor illuminance on brightness perception is almost negligible, but it violates common sense, which may be caused by the horizontal processing of indoor illuminance in the experiment being too close; ③ When the spatial geometry is not set up, the minimum per capita effective refuge area is 3.0 m2, and the minimum clear ceiling height is 2.0 m, the brightness perception data can be above 4.0; ④ When the spatial geometry is set up, the minimum clear ceiling height is 2.0 m, and the minimum per capita effective refuge area must reach 4.1 m2. When the minimum per capita effective refuge area is 3.0 m2, the minimum clear ceiling height must reach 3.0 m to ensure the brightness perception value is above 4.0.
Y5		Y5 = 0.330X3 + 0.228X4 + 0.338X6 + 0.289X9 + 1.859	(5)
	prediction	① Setting up a signage system and spatial geometry can improve the enjoyment of underground space; ② When the minimum width of the public passage is 1.5 m, the per capita effective refuge area must be at least 4.5 m2 to ensure that the Enjoyment is above 4.0; ③ When the minimum per capita effective refuge area is 3.0 m2, the width of the public passage must be at least 2.6 m to ensure that the enjoyment is above 4.0.
Y6		Y6 = 0.236X4 + 0.592X6 + 0.362X9 + 2.269	(6)
	prediction	① Setting up a signage system and spatial geometry can improve the index value of the spatial richness of underground space; ② When setting up a signage system and spatial geometry, the per capita effective refuge area must reach more than 3.3 m2 to ensure the spatial richness is above 4.0.
Y7		Y7 = 0.285X2 + 0.392X4 + 0.324	(7)
	prediction	①When the minimum per capita effective refuge area is 3.0 m2, the clear ceiling height must be at least 8.8 m to ensure that the sense of spatial openness is above 4.0; ②When the minimum clear ceiling height is 2.0 m, the minimum per capita effective refuge area must be at least 8.0 m2 to ensure that the sense of spatial openness is above 4.0.
Y8		Y8 = 0.908X2 + 0.097X4 + 0.666	(8)
	prediction	① When the minimum clear ceiling height is 2.0 m, the per capita effective refuge area must be at least 15.7 m2 to ensure that the sense of spatial openness is above 4.0; ② When the minimum per capita effective refuge area is 3.0 m2, the clear ceiling height is 3.4 m, which can ensure that the height perception value is above 4.0.
Y9		Y9 = 0.341X2 + 0.301X4—1.188X5 + 0.496X9 + 4.516 (The influence of X1 and X8 on Y9 is minimal, so they are ignored.)	(9)
	prediction	① When beds are used as accommodation, regardless of whether the signage system is set up or not, the minimum value of the clear ceiling height is 2.0 m, and the minimum value of the per capita effective refuge area is 3.0 m2, the index value of memorability can reach about 5.0; ② When tents are used as accommodation: A. If no signage system is set up when the minimum value of the clear ceiling height is 2.0 m, the per capita effective refuge area must reach 4.0 m2 or more to make the index value of memorability reach 4.0 or more; when the per capita effective refuge area is 3.0 m2 or more, the clear ceiling height must reach 2.8 m or more to make the index value of memorability reach 4.0 or more; B. If a signage system is set up, the minimum value of the clear ceiling height and the per capita effective refuge area can meet the requirements.
Y10		Y10 = 0.249X2 + 0.250X4 + 0.327X9 + 2.492	(10)
	prediction	① When the signage system is set up, the clear ceiling height is 2.0 m, and the per capita effective refuge area is 3.0 m2, which can make the index value of sense of order greater than 4.0; ② When the signage system is not set up, the clear ceiling height is 2 m–3.1 m, and the per capita effective refuge area is 3.0 m2 to 4.1 m2, which can meet the index value of sense of order greater than 4.0.
Y11		Y11 = 0.272X4 + 0.535X6 + 2.374	(11)
	prediction	①When the spatial geometry is set, the per capita effective refuge area is above 4.1 m2 to make the index value of perceived functionality above 4.0; ②When the spatial geometry is not set, the per capita effective refuge area is above 6.0 m2 to make the index value of perceived functionality above 4.0.

.

Consistent with the orthogonal experimental design, which prioritizes isolating main effects, the regression models focused on independent variable impacts. To address potential interaction effects (e.g., between X2 and X4, or X2 and X6), models incorporating interaction terms were subsequently tested. Results indicated no statistically significant interactions (all p > 0.05), confirming that within the study’s scope, main effects operate independently.

From the results of correlation analysis and multiple linear regression analysis, this research found that among the 11 spatial indicators, 7 spatial factors have a significant impact on people’s psychological indicators, namely clear ceiling height, public corridor width, per capita effective refuge area, accommodation layout, signage system, spatial geometry, and indoor illuminance. Among them, per capita effective refuge area has the most significant impact on psychological indicators, followed by clear ceiling height, signage system and spatial geometry, public corridor width, accommodation layout, and indoor illuminance.

The relationships between X and Y can be explained through theoretical mechanisms linking spatial factors to perceptual responses, as follows:

X4 (per capita effective refuge area), the most influential factor, impacts Y1 (Comfort), Y2 (Relaxation), Y3 (Affinity), and Y11 (Perceived Functionality) through “personal space preservation.” Sufficient per capita area reduces physical crowding and visual clutter, aligning with interpersonal distance norms to alleviate stress—directly enhancing comfort (Y1) and relaxation (Y2) (Equations (1) and (2)). It also ensures functional operability (e.g., movement, equipment access), boosting perceived functionality (Y11) (Equation (11)).

X2 (clear ceiling height) affects Y1 (0.208), Y7 (0.285), and Y8 (0.908) via “vertical spatial cognition.” Higher ceilings expand vertical visual fields, reducing confinement stress and activating associations with “openness,” which elevates comfort (Y1), spatial openness (Y7) (Equations (1) and (7), and height perception (Y8) (Equation (8)) by mapping physical dimensions to psychological evaluation.

X6 (spatial geometry) exerts dual effects: positive impacts on Y1 (0.312), Y6 (0.592), and Y11 (0.535) stem from “environmental complexity enhancement”—distinct geometric features (e.g., zoning) increase spatial richness, stimulating sensory engagement to boost comfort (Y1) and spatial richness (Y6) (Equations (1) and (6)).

X9 (signage system) influences Y1 (0.281), Y9 (0.496), and Y10 (0.327) through “cognitive mapping facilitation.” Clear signage reduces navigational ambiguity, lowering cognitive load to enhance relaxation (Y2) (Equation (2)), perceived order (Y10) (Equation (10)), and memorability (Y9) (Equation (9)) by aiding structured information encoding.

X3 (public corridor width) impacts Y5 (Enjoyment) (Equation (5): coefficient 0.330) via “spatial flow preservation.” Wider corridors (≥1.5 m) reduce congestion and movement restrictions, aligning with the psychological need for unobstructed mobility. This enhances hedonic experiences (Y5) by minimizing frustration from blocked pathways, as supported by predictions that corridor width ≥2.6 m (with 3.0 m² per capita area) ensures Y5 ≥ 4.0.

X5 (accommodation layout) affects Y9 (Memorability) (Equation (9): coefficient −1.188) through “spatial regularity perception.” Tents (as accommodation) introduce irregularity in layout, increasing cognitive ambiguity and reducing memorability (Y9), whereas beds (with fixed, structured placement) enhance spatial predictability—explaining why beds consistently yield higher Y9 scores (≥5.0) even with minimal space (3.0 m² per capita).

X8 (indoor illuminance) exerts a limited but non-negligible impact on Y4 (Brightness Perception), with its modest influence attributed to “narrow experimental variation.” While illuminance theoretically correlates with brightness perception, the minimal range of illuminance levels tested in this study (as noted in predictions) failed to generate substantial perceptual differences—explaining why its effect, though present, remains weaker compared to other factors like X2 or X4. This aligns with the observation that illuminance only marginally adjusts brightness perception under constrained experimental conditions.

However, the effective refuge area, interior color, greenery layout, and skylight do not significantly impact psychological indicators. For X1 (effective refuge area), tested at 1000 m², 2000 m², 3000 m², and 4000 m², statistical analysis confirmed no significant correlation with psychological indicators (all p > 0.05), with negligible differences in responses across these scales. This suggests that beyond 1000 m², spatial boundary perception weakens—large area sizes cease to act as salient psychological cues, reducing X1′s impact. Similarly, interior color only plays a supporting role and has less impact than the signage system. Greenery layout appears sparse due to a large number of beds or tents, or the two arrangements produce similar effects. At the same time, the skylight area is small, especially on the second underground floor and below, so the impact can be ignored.

Based on the combined prediction results, it can be concluded that among the seven factors, meeting the following conditions can lead to average scores above 4.0 across psychological indicators, indicating a generally acceptable psychological perception of underground refuge spaces: the clear ceiling height should range between 2.0 and 3.0 m; the public corridor width should be between 1.5 and 2.6 m; and the per capita effective refuge area should fall within 4.1–4.5 m². In addition, both beds and tents are acceptable accommodation types. It is recommended to include a signage system and a clearly defined spatial geometry, and to ensure that the average indoor illuminance exceeds 50 lx.

5. Discussion and Recommendations

Building on the questionnaire and VR experimental findings, this study revealed several psychological and behavioral tendencies toward underground refuge spaces. While instinctive perceptions remain predominantly negative, individuals—especially those with higher education levels—exhibit willingness to seek shelter underground when appropriately informed. The VR results further quantified how key spatial factors, including per capita effective refuge area, clear ceiling height, and signage systems, significantly affect users’ psychological comfort. These insights form the foundation for proposing specific design guidelines aimed at improving underground space usability during emergencies.

Compared with existing national shelter standards (GB51143-2015), this study highlights specific environmental parameters that can be better defined or optimized to integrate psychological comfort into technical specifications. While current guidelines focus primarily on minimum physical requirements (e.g., total area, fire zoning), our findings introduce psychologically derived thresholds that address user experience gaps. For instance, the recommended per capita effective refuge area (4.1–4.5 m²) exceeds the existing 3.0 m² standard to mitigate crowding stress, while the clear ceiling height range (2.0–3.0 m) and signage system specifications (e.g., density and placement) fill gaps in the current code, which lacks such details.

Table 12. Index comparison table.

Indicator	This Research	GB51143-2015 Code for Design of Disasters Mitigation Emergency Congregate Shelter
Effective Refuge Area	It is best to be less than 4000 m2	Accommodation Cluster ≤ 4320 m2, Accommodation Unit ≤ 1080 m2
Clear Ceiling Height	2.0–3.0 m	——
Public Corridor Width	1.5–2.6 m	1.5 m
Per Capita Effective Refuge Area	4.1–4.5 m2	3.0 m2
Accommodation Layout	combine tents with beds (with bed curtains)	Tent or portable housing
Spatial Geometry	Use elements such as roads, nodes, areas, landmarks, boundaries, etc., to arrange space	——
Interior Color	Use soothing colors or no color distinction	——
Indoor Illuminance	≥50 lx	≥50 lx
Signage System	Set up a signage system, and all kinds of signs are set up in the dormitory area	Fixed shelters should be equipped with area location indications, warning signs and site function demonstration signs
Greenery Layout	Both decentralized potted plants and centralized gardens can be considered	——
Skylight	Utilize existing skylight areas to create public spaces	——
Function settings	Health center, psychological counseling, canteen, open activity area, supermarket, library, gym, etc.	Emergency rest areas, public activity areas, emergency service points, public toilets, garbage collection points, material distribution points

summarizes these proposed thresholds as complementary enhancements to GB51143-2015. Specifically, our data support three avenues for future code updates: (1) incorporating psychological comfort metrics (e.g., stress-reducing spatial dimensions) alongside physical safety standards, including functional facilities tied to mental well-being—such as gyms and libraries (identified as key needs in the questionnaire Q19), where gyms alleviate confinement-induced stress and libraries foster relaxation (addressing concerns about engagement in the questionnaire Q14/Q15); (2) expanding parameters to include under-specified factors like spatial geometry and interior color, which our results link to perceived safety and relaxation; (3) refining existing requirements (e.g., signage systems) with user-centric details (e.g., placement intervals) to improve practical usability, while also integrating other user-identified functions (e.g., health centers, open activity areas from the questionnaire Q19) that mitigate anxiety and loneliness in prolonged refuge.

These additions would strengthen the code’s relevance to real-world user needs, ensuring refuge spaces are not only safe but also psychologically viable during prolonged occupancy.

Therefore, in the DUSR, it is recommended that real-world disaster prevention and refuge spaces be designed based on the principles revealed in this research:

(1)

Safety: Site selection should prioritize locations aligned with human behavioral habits, avoiding disaster sources and external risks. Operational guidance: Apply the 4.1–4.5 m² Per Capita effective refuge area threshold—use 4.5 m² in high-risk zones (e.g., near exits) for enhanced buffer, and 4.1 m² in low-risk core areas to optimize space efficiency.

(2)

Accessibility: Underground spaces should be zoned to ensure convenient access to refuge zones for all occupants. Operational guidance: Implement the 1.5–2.6 m public corridor width range: 2.0–2.6 m for main evacuation routes (high traffic) and 1.5–2.0 m for dormitory internal corridors (lower traffic). Ensure maximum 50 m walking distance from any area to the nearest refuge zone.

(3)

Identifiability: Enhance spatial recognition via distinct geometry and clear signage. Operational guidance: Combine spatial elements (roads, landmarks) with targeted signage placement—1 sign per 20 m in main corridors and 1 sign per 10 m in dormitory areas—to reinforce memorability.

(4)

Privacy: Control privacy through layout design, including gender separation and furniture arrangements. Operational guidance: Use tents for short-term stays (with movable partitions) and beds with ≥1.8 m curtains for long-term stays. Separate male/female areas with ≥2.2 m solid partitions (not just signage).

(5)

Comfort: Address vulnerable groups’ needs with barrier-free design and controlled spatial indicators. Operational guidance: Adjust clear ceiling height (2.0–3.0 m): 2.5–3.0 m in accessible routes, 2.0–2.5 m in standard areas. Set indoor illuminance to 50–75 lx for rest zones and 75–100 lx for health centers (exceeding the baseline 50 lx where critical).

6. Conclusions

Underground spaces, as key components of urban disaster prevention systems, possess inherent structural characteristics that offer significant advantages in disaster resistance. However, due to widespread psychological resistance in public perception, such spaces are often marginalized in emergency shelter decisions—even when their protective potential is objectively superior. Previous studies on the psychological impact of underground environments have primarily relied on qualitative methods, using phenomenological descriptions to reveal issues such as feelings of oppression and spatial disorientation experienced by users. In recent years, quantitative approaches have gained traction, typically following two main technical paths: the first employs statistical data to directly present patterns and perceptions, while the second applies mathematical models such as factor analysis to uncover deeper correlations. The former often proves inadequate when dealing with large-scale multidimensional data, as it struggles to reveal the internal relationships among influencing factors. The latter, though more insightful in theory, may suffer from limited sample sizes or incomplete survey scope.

To comprehensively and systematically examine the underground environment, this research selected a representative underground parking garage as the research object. Using both questionnaire surveys and VR-based experiments, it explores the underlying and direct relationships between human perception and spatial environment from two perspectives. Taken together, the main contributions of this research are as follows:

(1)

This research innovatively adopts environmental psychology and VR methods to investigate the DUSR, opening up new perspectives for the DUSR.

(2)

From the perspective of environmental psychology, this research employs questionnaire surveys to summarize patterns of refuge behavior, psychological response characteristics, and potential needs of people in underground environments, providing qualitative design references for the DUSR.

(3)

Using VR experimental methods, the research identifies the influencing factors of underground spaces on humans and their appropriate parameters, and analyzes the common factors underlying key psychological indicators.

(4)

This research compares the qualitative and quantitative indicators obtained from the research with current domestic disaster prevention and the Code (GB51143-2015), identifies existing limitations within these regulations, and evaluates the practical significance of the findings for the implementation of the DUSR in real-world contexts.

These findings offer practical implications for policy and planning. Emergency managers can integrate quantified psychological thresholds (e.g., brightness, spatial openness) into shelter design codes to improve public acceptance; designers may reference key influencing factors to balance safety and user experience; and disaster educators could adopt VR-based tools to enhance public familiarity with and compliance in using underground refuges during emergencies.

While these conclusions are widely applicable across urban contexts in China—given standardized underground garage designs and uniform emergency policies—their relevance in international settings requires further assessment. Cultural variations, such as differing levels of public trust in emergency institutions, preferences for personal autonomy, and risk perception, may shape behavioral responses in distinct regions. To address this, future research will focus on three directions to enhance policy relevance: exploring region-specific implementation strategies to align design parameters with diverse emergency management systems; developing inclusive adaptations for vulnerable groups (e.g., the elderly, people with disabilities) to ensure equitable access to psychologically supportive refuge environments; and conducting cross-cultural validation through comparative studies in other countries and institutional systems. These efforts aim to translate theoretical insights into globally applicable practical guidance.

However, this study has several limitations. First, the physical variables examined (X1–X11) focus on visual–spatial attributes (e.g., area, ceiling height, lighting), as these are theoretically grounded in environmental psychology [37] and technically manipulable in VR. Yet, due to current VR sensory limitations, factors like temperature, humidity, air quality, acoustic conditions, and thermal comfort—critical in real-world underground environments—could not be reliably simulated [68]. Ethical considerations (following APA guidelines [69]) also precluded simulating harmful conditions that might pose health risks. Second, reliance on short-duration VR experiences and semantic scale evaluations limits ecological validity, as these may not fully capture long-term user experiences or stress responses in realistic disaster scenarios. Third, despite covariate analyses confirming that age and vulnerability-related factors do not confound core outcomes (p > 0.05), the sample is skewed toward young adults (predominantly from urban Chinese contexts), with underrepresentation of the elderly, specific vulnerable groups (e.g., people with disabilities), and cultural diversity, which may limit generalizability across different sociocultural contexts.

To address these limitations, future research will (1) collaborate with practitioners to develop a multisensory underground simulation lab with controllable thermal, acoustic, and air-quality parameters to enhance VR realism; (2) incorporate wearable physiological monitoring to track longitudinal psychological responses under simulated emergency conditions; and (3) expand the participant pool to include a broader demographic range, improving the applicability of design recommendations.