Linking Human–Bird Interactions to Restorative Environmental Perception and Mental Health: A Landscape Perception Perspective

Abstract

Birds, as both wetland ecosystem health indicators and highly perceptible forms of wildlife, provide multi-sensory interaction opportunities shaping human health and well-being. However, most studies simplify birds into static landscape metrics, with limited attention to dynamic human–bird interactions and their mental health benefits. Grounded in landscape perception theory, this study constructs an “interaction–perception–restoration” framework and divides human–bird interactions into sensory, cognitive, and participatory levels based on cognitive resource investment. We collected 321 valid samples from Haizhu National Wetland Park. A mixed analytical strategy was adopted, using structural equation modeling to test the framework and moderated mediation models to examine differential pathways. The results showed the following: (1) Restorative environmental perception (REP) plays a partial mediating role between human–bird interactions and mental health, explaining 46.17% of the total effect. (2) All three interaction levels significantly enhance mental health, with cognitive interaction showing the strongest direct effect (β = 0.347 ***) and sensory interaction the largest indirect effect through REP (β = 0.194 ***). (3) Environmental characteristics directly improve REP (β = 0.51 ***) but do not significantly moderate the relationship between human–bird interactions and REP. This study highlights interaction quality and depth as core drivers of mental health, offering insights for optimizing ecological and recreational services in urban wetland parks.

1. Introduction

With the acceleration of global urbanization, urban residents are facing issues such as reduced natural contact opportunities and increased cognitive load [1]. How to optimize the experience of urban green spaces to promote attention restoration and relieve stress has become a key topic in recent research [2,3]. As an important component of urban green space, wetland parks serve not only as ecological infrastructures that safeguard biodiversity but also as crucial recreational landscapes where residents can reconnect with nature and seek psychological restoration [4,5]. Within wetland parks, birds, as both health indicator species of wetland ecosystems [6] and highly perceptible forms of wildlife for human appreciation, serve as excellent subjects for research. Their appearance and activities greatly enrich urban residents’ natural experiences in wetland parks [7], while they are considered to be associated with human well-being [8,9,10]. Studies have shown that bird-watching or bird sounds can effectively stimulate positive emotion, promote psychological recovery, and relieve stress [11,12,13]. Understanding and respecting both the ecological and cultural values of birds is essential for protecting and utilizing wetland resources, ultimately promoting a harmonious coexistence between humans and nature.

However, existing studies have mostly focused on the effects of static landscape metrics such as bird presence [14,15], bird diversity [16,17,18,19,20], bird richness [21,22], and bird sounds [23,24,25,26] on mental health. They have also explained the mediating role of restorative environmental perception (REP) from the perspective of attention restoration theory (ART; [27]) and stress reduction theory (SRT; [28]). However, these studies mostly focus on cognitive outcome levels and fail to fully attend to the interaction process itself [29], ultimately leading to insufficient understanding of the mechanisms by which human–bird interactions influence mental health. In fact, human–bird interactions are not merely passive nature exposure, but include multiple behavioral levels, including sensory reception (e.g., hearing, vision; [30]), cognitive processing (e.g., species identification, behavior understanding; [12]), and active participation (e.g., bird rescue, citizen science; [14]). Different behavioral levels involve varying cognitive resource investment and emotional activation levels, which may correspond to different mental health benefits. Therefore, it is necessary to explore the differentiated impact of human–bird interaction depth on psychological recovery from the internal structure of behaviors.

Moreover, human–bird interactions do not occur in isolation but within specific environmental contexts [31]. The environmental characteristics of wetland parks, such as landscape diversity, visual transparency, facility practicality, etc., will not only affect people’s subjective evaluation of the naturalness, beauty, and safety of the place [32,33]. These characteristics may also affect the quality of human–bird interactions, which in turn modulate the impact of interaction on psychological recovery. Although existing studies have explored the impact of environmental characteristics on mental health [34,35,36,37], few studies have combined them with specific human–bird interactions to investigate the moderating role of environmental characteristics in the process of human–bird interactions affecting mental health.

In summary, this study is based on landscape perception theory, taking urban wetland parks as the research object. Its core innovations lie in three aspects that advance existing theory: first, by decomposing human–bird interactions into sensory, cognitive, and participatory levels based on cognitive resource investment, it breaks through the limitation of treating birds as static landscape metrics in existing studies and advances the refinement of dynamic human–bird interaction research; second, it constructs an “interaction–perception–restoration” theoretical framework to systematically reveal the direct and indirect impact paths of human–bird interactions on mental health, extending the mediation theory to the field of dynamic human–bird interactions; third, it explores the moderated role of environmental characteristics in the above mechanism, enriching the situational implications of existing interaction theories. The study aims to answer the following specific research questions:

(1)

What are the differences in the intensity and paths of the impacts of different levels of human–bird interactions on mental health?

(2)

Does REP play a mediating role between human–bird interactions and mental health, and are there differences in the mediating effects among different interaction levels?

(3)

Do environmental characteristics moderate the relationship between human–bird interactions and REP, and is there heterogeneity in the moderating effects across different interaction levels?

This study attempts to construct an explanatory framework from three aspects: interaction level, perception mechanism, and environmental context. It provides empirical basis for understanding the mental health benefits of dynamic human–bird interactions and offers theoretical support for the design and management strategies of urban wetland parks to optimize ecological and recreational services.

2. Literature Review and Hypothesis Development

2.1. Urban Wetland Parks, Human–Bird Interactions, and Mental Health

Urban green spaces, including parks, gardens, street greenery, etc., are widely recognized as important factors for enhancing the mental health and human well-being [2,3,34]. Green spaces provide opportunities for natural contact, helping people escape urban hustle and bustle and relax both physically and mentally, thereby effectively relieving stress, restoring attention, and enhancing positive emotion [15,19]. In urban wetland parks, a unique type of green space, opportunities for human–bird interactions are more prominent. Urban wetland parks are defined as parks that possess wetland ecological functions and typical characteristics, playing a variety of functions such as ecological conservation, science popularization and education, leisure, and recreation [7]. Compared with general urban green spaces, wetland parks typically show higher biodiversity, especially bird diversity, due to their rich natural water, wetland vegetation, and wildlife resources [4]. These high-ecological-value habitats provide urban residents with rich opportunities for contact with birds and have great potential for mental health promotion. However, while existing studies on the relationship between urban green spaces and mental health have mentioned the role of birds, research specifically targeting wetland parks where opportunities for human–bird interaction are particularly prominent remains limited. This fails to fully explore the unique value of bird-related experiences in this distinct green space type.

Birds, as one of the most common and easily perceptible wildlife in wetland parks, are increasingly becoming a focus in research on natural experiences. Numerous empirical studies have shown that the appearance and activities of birds can have positive effects on human mental health. For example, bird-watching has been shown to improve subjective well-being and reduce psychological distress among college students [10], and to help maintain cognitive resources, activity capacity, and bio-psycho-social health in older adults in nursing homes [13]. Further studies have explored the impact of bird diversity on mental health. Higher bird diversity near residences has been linked to lower depression levels and better mental health scores [20], with Chen, et al. [21] further confirming the mediating role of bird richness on mental health in green spaces. In addition, bird sounds also have significant mental health value. Ratcliffe, et al. [11] identified them as the natural sounds most strongly associated with stress recovery and attention restoration, while Jahani, et al. [24] used artificial intelligence to demonstrate that specific combinations of bird sounds can further enhance psychological recovery. Although these studies fully reveal the effects of bird presence, diversity, richness, and sounds on mental health, most of them regard birds as static landscape metrics, ignoring the initiative and complexity of human–bird interaction processes. In fact, the interactions between humans and birds are not one-way environmental exposure but dynamic processes that include multiple behaviors such as observation, listening, identification, monitoring, and rescue. Such interactions may have deeper emotional and cognitive value than passive contact, and the mechanisms for promoting mental health may be also different [38].

2.2. Different Level of Human–Bird Interactions

Human–animal interaction (HAI) is an important component of the relationship between humans and nature [39,40]. In urban areas, HAI can not only lead to positive experiences but also have negative impacts [41] and even cause conflicts [42]. This study focuses on the positive interactions between humans and birds in urban wetland parks, dividing them into different levels based on the depth and complexity of the interaction. Based on the representative theory of landscape perception [28,43,44], Zigmunde, et al. [45] proposed a multi-perception interaction framework, arguing that human perception needs to go through the cyclic operation of three major modules—visual perception, sensory perception, and cognitive perception—playing a decisive role in the arousal of human cognition and behavior. The skills–rules–knowledge (SRK) framework proposed by Vicente and Rasmussen [46] divides human behavior into three levels according to the degree of cognitive resource investment. These three levels are skill-based level (unconscious automatic responses), rule-based level (rule application and judgmental reasoning), and knowledge-based level (problem solving and decision making). Based on these, this study classifies human–bird interactions into three levels, sensory interaction, cognitive interaction, and participatory interaction, corresponding to cognitive processing and behavioral complexity from low to high.

Specifically, sensory interaction mainly refers to the process by which people passively receive information about birds through their senses [47]. For example, observing the morphology, color, or behavior of birds through vision [21] and perceiving the sounds of birds through hearing [23,24,48]. In addition, Urry [49] proposed an atypical sense called “kinesthesia”, which informs the body’s actions in space through the sensations of movement registered in its joints, muscles, tendons, etc. For instance, turning the neck, stopping to observe, crouching down, or even moving slightly through branches to get a better view. Since these kinesthetic behaviors are largely unconscious and serve to reinforce sensory processes, they are categorized under the sensory interaction level [30].

Furthermore, cognitive interaction is the process by which people selectively process and actively understand bird-related information based on sensory interactions. It includes identifying bird species, inferring their habits, analyzing the characteristics of their environment, or linking current observations with past knowledge and experience [15,31]. This level involves higher-level cognitive processes such as concentration, information integration, and conceptual construction. People often use bird identification guides, bird–watching apps, or discussions with peers to supplement and correct their cognition, which can reflect differences in their ecological cognitive abilities and sense of connection to nature [8,9]. The subject of cognitive interaction is human beings, which determines the possibility of forming meaningful connections between humans and birds.

Moreover, participatory interaction is the highest level of human–bird interaction, expressed as people actively participating in bird-related organized activities, such as bird-watching competitions, nature education classes, bird photography, species monitoring and rescue, etc. [14,50]. Such interactions usually require investment of time, energy, and resources, and rely on a level of knowledge, skills, or group collaboration. Compared with the previous two types of interactions, participatory interaction is more action-oriented and emotionally engaged. Its purpose is not limited to information reception or understanding, but rather to experience, contribute to, and even change natural systems [51,52]. This level of interaction tends to enhance people’s environmental responsibility, but is accompanied by higher cognitive resource requirements. The specific mechanism of its influence on mental health requires further exploration.

Based on the above analysis, this study suggests that different levels of human–bird interactions may exert varying impacts on people’s mental health due to differences in cognitive resource investment and behavioral depth. Although existing studies have noted many human–bird interactions, there remains a lack of a clear hierarchical framework to systematically compare their mental health benefits. As interaction levels increase, emotional connections between humans and birds may strengthen, thereby triggering deeper psychological recovery benefits. However, excessive cognitive load may bring resource consumption and psychological burden, potentially creating a threshold effect between cognitive load and psychological recovery. Thus, the following hypotheses are proposed:

H1: 

Human–bird interactions have significant positive impacts on mental health.

H1a: The effect intensity of different levels of human–bird interactions on mental health differs.

2.3. The Mediating Role of Restorative Environmental Perception

ART and SRT are two classical theories to explain the benefits of the natural environment on human mental health. ART emphasizes the ability of natural environments to recover overused cognitive resources, especially directed attention, through the property of non-coercive attention attraction (soft fascination; [27]), while SRT emphasizes that natural exposure can rapidly reduce physiological arousal levels, alleviate negative emotions and stress responses, and thus facilitate positive emotional shifts [28]. Although the theoretical starting point is different, both theories emphasize the critical role of the personal subjective perception of environmental restoration attributes.

Based on this, REP is regarded as a mediating variable linking natural environments and mental health. REP refers to personal subjective perception of whether a particular environment provides restorative experiences, typically encompassing four dimensions: (1) being-away: distance from sources of everyday stress and anxiety; (2) fascination: possessing captivating attributes; (3) compatibility: alignment with personal expectations and preferences; and (4) extent: richness and continuity of the landscape [53]. REP has been shown to mediate the relationship between natural exposure and mental health (e.g., positive emotions, well-being, and vitality [3,19,30]).

In the process of human–bird interactions, different levels of interactions may indirectly affect emotions and well-being by stimulating humans to develop REP. For example, (1) sensory interaction: melodious bird sounds or colorful birds can enhance the attractiveness of the environment and help people create a sense of distance from everyday distractions; (2) cognitive interaction: identifying specific bird species and understanding bird habits can make the environment more meaningful and understandable, increasing the perception of environmental compatibility and continuity; and (3) participatory interaction: active participation in activities such as bird–watching, volunteering, or training, which usually requires investment of substantial time and effort. They may significantly enhance people’s feelings of being-away and of compatibility, and deepen emotional connection and belonging to the environment. These processes collectively enhance people’s restorative perception of wetland parks, thereby improving emotional state and subjective well-being.

It should be noted that the pathway of human–bird interactions on mental health does not depend entirely on REP. Some studies have shown that direct interactions with animals can immediately trigger pleasant experiences and emotional regulation responses, indicating the existence of a direct pathway without cognitive mediation [54]. However, in most situations, REP remains an important cognitive mechanism that plays a significant role in explaining how interactive behavior translates into psychological benefits. Therefore, based on ART, SRT, and existing research on the mediation of REP, this study proposes the following hypotheses:

H2: 

REP plays a mediating role between human–bird interactions and mental health.

H2a: Human–bird interactions have a significantly positive effect on REP.

H2b: REP has a significant positive effect on mental health.

H2c: The mediating effect intensity of REP differs across the relationships between different levels of human–bird interactions and mental health.

2.4. The Moderating Role of Environmental Characteristics

Human–bird interactions and the psychological benefits they generate do not exist in isolation from their environment. The environmental characteristics of urban wetland parks will deeply affect interaction behaviors and the effects of restorative experiences. Studies have shown that high-quality natural environments enhance the perceived quality of natural elements, thereby promoting attention restoration and emotional improvement [2,34,35].

From the perspective of environmental perception and restorative experiences, this study selects visual features, soundscape features, and facility features [55,56] as moderating variables through research and screening of the literature. We explore their potential moderating effects on the path from human–bird interactions to REP: (1) visual features: determine whether people can easily discover and observe birds (e.g., landscape diversity, visual openness, color richness), which directly affects the attractiveness of sensory interaction and the coherence of cognitive interaction [35,36]; (2) soundscape features: such as birdsong purity and noise interference, which can enhance or weaken restorative experiences at the auditory level and are important conditions affecting the quality of sensory interaction [37]; and (3) facility features: including accessibility, bird–watching platforms, and navigation systems, which influence the convenience and depth of visitor participation behaviors and are important supports for cognitive interaction and participatory interaction [4,30].

These environmental characteristics may not only enhance perceptual stimulation and behavioral convenience during interactions but also affect people’s subjective evaluations of environmental restorativeness. For example, noisy or visually restricted environments may weaken the effectiveness of sensory interaction, while well-equipped, pleasant soundscape environments may help stimulate deeper cognitive or participatory interaction [37,57]. Thus, this study proposes the following hypotheses:

H3: 

Environmental characteristics moderate the relationship between human–bird interactions and REP.

H3a: The moderating effect of environmental characteristics differs across the relationships between different levels of human–bird interactions and REP.

2.5. Theoretical Model Construction

Existing studies on human–bird interaction and mental health can be categorized into three strands. First, research on static landscape metrics has confirmed that bird presence [14,15], bird diversity [16,17,18,19,20], bird richness [21,22], and bird sounds [23,24,25,26] are positively correlated with mental health restoration. Second, a growing body of literature has shifted to dynamic interactions, showing that activities like bird–watching [15,31,58] and nature education [14,50] enhance well-being, but these studies rarely decompose interaction into distinct levels or compare their differential effects. Third, REP is recognized as a critical mediator [3,19,30], yet few studies have linked it to hierarchical human–bird interactions. Notably, the moderating role of environmental characteristics remains underexplored. The existing work either ignores context [59] or treats environments as homogeneous [15,20]. These gaps highlight the need for a framework that systematically examines multi-level interactions, their mediating mechanisms via REP, and contextual moderators.

Landscape perception theory [28,43,44], from the perspective of environmental psychology, discusses how individuals receive natural environmental stimuli through multiple senses (such as vision, hearing, etc.) and process them cognitively to form an understanding and construct meaning from these stimuli, thus triggering corresponding emotional and behavioral responses. Bell [60] pointed out that environmental perception is not only passive reception of external stimuli but also an active psychological process in which individuals process and integrate information based on their cognitive resources.

Based on the above theoretical perspectives and hypothetical logic, this study constructs a theoretical model of “interaction–perception–restoration” to explain the mechanism of human–bird interactions on mental health in urban wetland parks. The model takes human–bird interaction (multi-sensory and cognitive processing processes) as the starting point, regards REP (subjective perception of environmental restorative attributes) as the mediating link, and finally points to mental health (emotional response outcome), which is completely consistent with the core logic of landscape perception theory. It emphasizes the initiative of individuals in the interaction process and the transformation path from environmental interaction to psychological benefits through perception. It also introduces environmental characteristics as moderating variables. The study explores the pathway mechanism of human–bird interactions on mental health benefits for urban residents in different situations (Figure 1).

3. Materials and Methods

This study adopted a mixed analytical strategy to investigate the “interaction–perception–restoration” model. Figure 2 outlines the workflow developed for our research. Initially, a structured questionnaire was developed to collect self-reported data. This instrument was constructed by adapting established scales to measure the core constructs of multi-level human–bird interactions, environmental characteristics, restorative environmental perception (REP), and mental health. Subsequently, various analyses were conducted to test the hypotheses and address the research questions. These included (1) reliability and confirmatory factor analysis (CFA) in SPSS 27.0 and AMOS 26.0 to assess the consistency and validity of the measurement scales; (2) structural equation modeling (SEM) in AMOS 26.0 to test the overall theoretical model and the primary mediating effect of REP; and (3) a series of moderated mediation analyses using PROCESS v4.1 in SPSS 27.0 to disentangle the distinct pathway effects of different interaction levels and to examine the potential moderating role of environmental characteristics.

3.1. Study Areas

This study takes the Haizhu National Wetland Park in Guangzhou, China, as a case study. Located in the central urban area of Guangzhou, the park covers an area of about 1100 hectares. It is one of the largest and most well-preserved urban central wetlands in the world. It also serves as an important stopover place for international migratory birds and is a key ecological node on the migration corridor of waterbirds in the Pearl River Delta [21]. The wetland park not only exhibits typical wetland ecosystem characteristics, but also undertakes multiple functions, including urban ecological restoration and public education and leisure. Since the main purpose of this wetland park is ecological protection, a large part of its area is inaccessible to the public [61]. However, this study focuses on human–bird interactions, so the selected area should be accessible to the general public. Therefore, this study selected the main public open areas of this park as the case study site, mainly covering 4 wetland bird–watching trails in Haizhu Lake and the second phase of the park (Figure 3). These trails connect the natural habitats (such as lakes, sparse forest grasslands, and grass beaches) in the wetland park and intersperse bird–watching spaces (such as platforms, huts, and ecological exhibition halls), providing an ideal environment for studying the mental health benefits of human–bird interactions in urban wetland parks.

3.2. Variable Measurement and Scale Construction

This study was conducted through a combination of online and offline questionnaire surveys. The online questionnaire consisted of five parts. The questionnaire content was validated through expert consultation and pre-tested by 10 target participants for comprehensibility. In order to ensure the semantic clarity and contextual adaptability of the scale items, this study adopted the behavior frequency scale with contextual anchors [62,63], improving the validity and understandability of measurements by specifying bird–watching behavior scenarios. During the questionnaire development process, the scale was optimized based on interviews with two senior birders with more than three years of bird–watching experience, thereby enhancing the content validity and practicality of the questionnaire.

The first part collects demographic characteristics and bird–watching behavior profiles of the respondents, including gender, age, education level, and identity. “frequency”, “duration”, and “experience” are used as indicators to measure the intensity of human–bird interactions, describing the extent and persistence of individual contact with birds [10]. The second part measures human–bird interactions in wetland parks. According to the previous theoretical analysis, human–bird interaction measurement is divided into three sub-dimensions: sensory interaction, cognitive interaction, and participatory interaction. The scale was developed using the findings from Rutter, et al. [14] and Chen and Chen [51], and was localized to reflect common human–bird interaction scenarios in wetland parks, resulting in a final scale with 10 items. The third part evaluates the subjective perception of individuals on the environmental characteristics of wetland parks, covering three sub-dimensions of visual features, soundscape features, and facility features. It references the environmental experience dimensions proposed by Grahn and Stigsdotter [64] and combines them with the local context of this study, including 6 items in total. The fourth part measures restorative environmental perception (REP) of human–bird interactions. The PRS scale, developed based on the attention restoration theory (ART), was revised by Liu, et al. [65] and Wang, et al. [66] to adapt it to the Chinese urban park context. Further modifications were made to suit the specific context of this study, resulting in an 11-item scale covering four dimensions: being-away, fascination, compatibility, and extent. The fifth part measures three dimensions of individual mental health status, including positive emotion, negative emotion, and subjective well-being. Among them, emotional indicators were selected from typical emotional items in the positive and negative affective scale (PANAS; [67]), while well-being measurement was based on the world health organization five-item well-being index (WHO-5; [68]), with a total of 15 items.

The pretest mainly improved four aspects: (1) Comprehensibility: 8 participants reported ambiguity in items like “I closely observed the birds”, which was revised to “I carefully observe bird features such as plumage, posture, or patterns” to specify scenarios. (2) Response efficiency: 6 participants noted redundancy in 2 reverse-scored items, which were simplified to shorten completion time to 6–10 min. (3) Practicality: two senior bird–watchers (with over 3 years of experience) were interviewed separately but both suggested the necessity of providing explanations for professional terms such as citizen science. (4) Operability: three experts (landscape ecology professor, animal behavior expert, senior bird–watching guide) reviewed the questionnaire. They confirmed alignment with theories and the full coverage of core variables, with only minor tweaks (e.g., “environmental comfort” change into “visual openness” and “sound purity”) to boost operability.

The final questionnaire consists of 44 items, as shown in

Table A1. Measurement scale for human–bird interaction in urban wetland parks.

Scale	Dimension	Operational Definition	Item
Human–Bird Interaction	Sensory Interaction	Vision	I carefully observe bird features such as plumage, posture, or patterns.
			I often notice special bird behaviors, such as foraging, courtship, nesting, or flocking.
		Sound	I usually pay attention to bird sounds and even try to locate the source of the sounds.
		Kinesthesia	I usually adjust my posture (e.g., turning, crouching) to gain a clearer view of birds.
	Cognitive Interaction	Tool Learning	I learn to use binoculars, field guides, or bird–watching apps to observe details.
		Information Understanding	I seek information or ask others to better understand bird behaviors.
		Identification Ability	I learn to identify common bird species and understand their habitat preferences.
	Participatory Interaction	Educational Activities	I have attended nature classes, bird festivals, or guided bird–watching events.
		Volunteering	I have participated in bird-related volunteering, such as habitat maintenance or rescue.
		Citizen Science	I have uploaded records to birding apps or joined surveys contributing scientific data.
Environmental Characteristics	Visual Features	Landscape Diversity	The area has diverse and layered natural landscapes (wetlands, forests, grasslands) suitable for various bird species.
		Visual Openness	Bird–watching spots have open views, unobstructed by tall grass, trees, or crowds.
		Color Richness	The overall color coordination of water, plants, and birds is rich and harmonious.
	Soundscape Features	Sound Purity	I can clearly hear natural sounds, such as birdsong, wind, and water, without being masked by noise.
	Facility Features	Accessibility	Park trails, bird–watching facilities, and guidance systems meet the needs of elderly, children, and people with disabilities.
		Infrastructure	I can easily access bird–watching tools, such as information boards, binocular rentals, or guidebooks.
Restorative Environmental Perception	Being-Away	Isolation from Urban Noise	Birdsong and natural sounds help me temporarily escape city noise, feeling relaxed and peaceful.
		Attention Shift	I become absorbed in observing birds and temporarily forget daily stress and worries.
	Fascination	Visual Attraction	Colors and lively movements of birds provide visual appeal and enjoyment.
		Interest Attraction	Identifying birds, or engaging in gamified interactions stimulates curiosity and enjoyment.
	Compatibility	Sensory Comfort	Lighting, temperature, and scents in bird–watching places make me feel comfortable.
		Rhythm Match	The pace (speed and duration) of bird–watching activities suits my physical strength and preferences.
		Information Match	The bird knowledge provided by the science popularization sign is of moderate difficulty and meets my knowledge needs.
		Social Flexibility	I can freely choose to be alone or interact with others during bird–watching.
	Extent	Sensory Coherence	Bird–watching routes are well-designed, with distinct yet smoothly transitioning experiences.
		Cognitive Coherence	Bird–watching aids (signs, guides) are closely linked to the environment, helping me build ecological understanding.
		Participatory Coherence	Bird–watching activities (guided tours, classes, volunteering) are clear and accessible, supporting gradual learning.
Mental Health	Positive Emotions (At the time)	Interested	I feel interested when I encounter a rare bird or observe unusual behaviors.
		Excited	I feel excited when I discover a new bird species or hear a beautiful bird song.
		Proud	I feel proud when I gain valuable bird–watching experiences or learn new knowledge.
		Inspired	I feel inspired when observing birds deepens my appreciation of nature.
		Attentive	I feel attentive when identifying bird species or record their behaviors.
	Negative Emotions (At the time)	Distressed	I feel distressed when I fail to find the target bird species or when equipment malfunctions.
		Upset	I feel upset when distracting noises interrupt my bird–watching.
		Scared	I feel scared when sudden weather changes or noise disrupts my bird–watching.
		Jittery	I feel jittery when waiting a long time for birds to appear or missing the best time to observe them.
		Afraid	I feel afraid when the environment is too quiet or hearing unexpected sounds.
	Well-Being (Two weeks before and after)	Cheerful and Energetic	I have felt cheerful and full of energy when planning or engaging in bird–watching.
		Calm and Relaxed	I have felt calm and relaxed when listening to birdsong or after bird–watching.
		Active and Alert	I have felt active and mentally alert when identifying bird species or interpreting their behaviors.
		Life is Meaningful	I have felt that my life is more meaningful when I appreciate the vitality and freedom of birds.
		Connected with Others	I have felt emotionally connected when sharing bird–watching experiences with others.

. The items in the WHO-5 scale are scored from 0 to 5, while all other scales use a 5-point Likert scale (1 = “strongly disagree”, 5 = “strongly agree”). We conducted a multicollinearity test on all relevant independent variables associated with human–bird interaction and environmental characteristics, with mental health as the dependent variable. The results (

Table A2. Results of variance inflation factor (VIF) test.

Scale	Dimension	Independent Variable	1/VIF	VIF
Human–Bird Interaction	Sensory Interaction	Vision	0.602	1.661
			0.452	2.212
		Sound	0.515	1.943
		Kinesthesia	0.526	1.901
	Cognitive Interaction	Tool Learning	0.360	2.781
		Information Understanding	0.343	2.916
		Identification Ability	0.321	3.116
	Participatory Interaction	Educational Activities	0.371	2.696
		Volunteering	0.403	2.483
		Citizen Science	0.502	1.993
Environmental Characteristics	Visual Features	Landscape Diversity	0.663	1.508
		Visual Openness	0.704	1.420
		Color Richness	0.681	1.469
	Soundscape Features	Sound Purity	0.674	1.484
	Facility Features	Accessibility	0.731	1.369
		Infrastructure	0.749	1.336

Dependent variable: mental health.

) showed that the Variance Inflation Factor (VIF) values of all independent variables are below the critical threshold of 5, and their Tolerance values are greater than 0.1, indicating no severe multicollinearity. The reliability (Cronbach’s α > 0.7) of each dimension is good, making it suitable for structural equation modeling analysis.

3.3. Data Collection

The formal study was conducted from 10 April 2025, to 16 July 2025, simultaneously using an online questionnaire platform and field surveys in urban wetland parks. The online survey was hosted on the Tencent questionnaire platform, a widely used platform in China. Survey links were distributed via WeChat to target groups, including wetland park tourists, park volunteers, bird–watchers, nature education practitioners, and relevant scholars. Participants were required to have contact with birds (e.g., observation, listening, or participation in related activities) at least once in the wetland park in the past three months. A total of 324 questionnaires were collected in this survey. They were rigorously screened based on criteria such as the rationality of the questionnaire filling time (6–10 min), consistency of reverse questions, and regularity of data options. A total of 42 invalid questionnaires were deleted, resulting in 282 valid questionnaires, an effective recovery rate of 87.04%. In addition, field surveys were carried out in the Haizhu National Wetland Park three times, covering weekdays, legal holidays, and weekends (10 April 2025, 1 May 2025, and 17 May 2025, respectively), by randomly distributing paper questionnaires to visitors. A total of 48 questionnaires were collected from the field, with 9 invalid questionnaires excluded, resulting in 39 valid questionnaires, an effective response rate of 81.25%.

To minimize the possibility of a single survey method bias, reduce information waste, and effectively utilize data resources, the data collected online and offline were merged for analysis. According to the research by Lin, et al. [30], before merging the data, an independent sample t-test was performed to assess the overall consistency between the two groups of samples from different sources, ensuring that they could be considered from the same population [69,70]. The test results showed no significant differences (all p values > 0.05) between the two groups of samples for human–bird interaction (p = 0.199), environmental characteristics (p = 0.53), restorative environmental perception (p = 0.052), and mental health (p = 0.07), indicating preliminary consistency for potential merging.

To further address the concern about the small offline sample size (n = 39) and validate data merging rationality, we supplemented comprehensive comparative analyses and sensitivity analyses with detailed results in Appendix A,

Table A4. Comparison of descriptive statistics for core variables in SPSS 27.0.

Variable	Metrics	Total Sample (n = 321)	Weighted Merged Sample (n = 253)	Online Sample (n = 282)	Offline Sample (n= 39)
Human–Bird Interaction	Mean	2.7530	2.7736	2.7773	2.5769
	Standard Deviation	0.9126	0.9122	0.9118	0.9103
	Skewness	0.186	0.156	0.151	0.470
	Kurtosis	−1.139	−1.155	−1.158	−0.818
Environmental Characteristics	Mean	3.5272	3.5325	3.5335	3.4815
	Standard Deviation	0.4837	0.4899	0.4908	0.4321
	Skewness	0.092	0.081	0.079	0.136
	Kurtosis	−0.071	−0.073	−0.077	−0.079
REP	Mean	3.9550	3.9759	3.9797	3.7762
	Standard Deviation	0.6134	0.6071	0.6057	0.6462
	Skewness	−0.221	−0.184	−0.176	−0.428
	Kurtosis	−0.106	−0.277	−0.314	0.972
Mental Health	Mean	4.3542	4.3789	4.3833	4.1436
	Standard Deviation	0.7753	0.7589	0.7556	0.8882
	Skewness	−0.635	−0.581	−0.569	−0.805
	Kurtosis	1.016	0.938	0.912	1.077

Weighted Merged Sample (n = 253; weighted by the proportion of the online sample (n = 282) to the total sample (n = 321) and the offline sample (n = 39) to the total sample (n = 321), aiming to balance the size gap between online and offline samples)

,

Table A5. Comparison of variable Pearson correlation analysis in SPSS 27.0.

Variable Pairs	Total Sample (n = 321)	Weighted Merged Sample (n = 253)	Online Sample (n = 282)	Offline Sample (n = 39)	Consistency Level (Range of Differences)
Human–Bird Interaction– Environmental Characteristics	0.091	0.073	0.07	0.248	Minor Difference (5–10%)
Human–Bird Interaction–REP	0.368 **	0.367 **	0.367 **	0.342 *	Minimal Difference (0–5%)
Human–Bird Interaction–Mental Health	0.533 **	0.539 **	0.54 **	0.477 **	Minimal Difference (0–5%)
Environmental Characteristics–REP	0.561 **	0.553 **	0.552 **	0.646 **	Minor Difference (5–10%)
Environmental Characteristics–Mental Health	0.37 **	0.351 **	0.348 **	0.534 **	Minor Difference (5–10%)
REP–Mental Health	0.649 **	0.62 **	0.615 **	0.83 **	Moderate Difference (10–20%)

** means p < 0.01; * means p < 0.05.

,

Table A6. Comparison of mediation analysis results using PROCESS v4.1 in SPSS 27.0.

Mediation Model Indicators	Total Sample (n = 321)	Online Sample (n = 282)	Offline Sample (n = 39)	Consistency Level (Range of Differences)
Human–Bird Interaction → REP	0.2476 ***	0.2438 ***	0.2425 *	Minimal Difference (0–5%)
REP → Mental Health	0.6621 ***	0.6004 ***	1.0370 ***	Moderate Difference (10–20%)
Direct Effect Intensity	0.2893 ***	0.3008 ***	0.2144 *	Minor Difference (5–10%)
Indirect Effect Intensity	0.1639 ***	0.1464 ***	0.2514 **	Moderate Difference (10–20%)
Total Effect Intensity	0.4532 ***	0.4472 ***	0.4658 **	Minimal Difference (0–5%)

*** means p < 0.001; ** means p < 0.01; * means p < 0.05.

,

Table A7. Comparison of reliability, validity, and discriminant validity in confirmatory factor analysis (CFA) in AMOS 26.0.

Variable	Item	Metrics	Total Sample (n = 321)	Online Sample (n = 282)	Degree of Difference
Human–Bird Interaction	Sensory Interaction	Factor Loading	0.855	0.859	0.47%
	Cognitive Interaction	Factor Loading	0.961	0.96	0.1%
	Participatory Interaction	Factor Loading	0.694	0.687	1%
	/	Cronbach’s α	0.917	0.916	0.1%
	/	AVE	0.712	0.711	0.14%
	/	CR	0.879	0.879	0%
Environmental Characteristics	Visual Features	Factor Loading	0.763	0.759	0.52%
	Soundscape Features	Factor Loading	0.653	0.68	4.13%
	Facility Features	Factor Loading	0.528	0.528	0%
	/	Cronbach’s α	0.722	0.724	0.28%
	/	AVE	0.429	0.439	2.33%
	/	CR	0.688	0.697	1.3%
REP	Being-Away	Factor Loading	0.725	0.738	1.79%
	Fascination	Factor Loading	0.74	0.742	0.27%
	Compatibility	Factor Loading	0.812	0.796	1.97%
	Extent	Factor Loading	0.749	0.759	1.33%
	/	Cronbach’s α	0.889	0.89	0.11%
	/	AVE	0.573	0.576	0.52%
	/	CR	0.843	0.845	0.41%
Mental Health	Positive Emotion	Factor Loading	0.848	0.851	0.35%
	Well-Being	Factor Loading	0.871	0.867	0.46%
	/	Cronbach’s α	0.918	0.915	0.33%
	/	AVE	0.739	0.738	0.14%
	/	CR	0.85	0.849	0.12%
All Constructs	/	Discriminant Validity	Reach the Standard	Reach the Standard	/

Degree of Difference = |Total Sample Value − Online Sample Value|/Total Sample Value × 100%.

and

Table A8. Comparison of core indicators in structural equation model (SEM) in AMOS 26.0.

SEM Core Indicators	Total Sample (n = 321)	Online Sample (n = 282)	Degree of Difference	Consistency Level (Range of Differences)
Model Fit Indices	χ2/df = 2.938, RMSEA = 0.076, CFI = 0.957	χ2/df = 2.552, RMSEA = 0.074, CFI = 0.96	/	All Indices Meet Standards (χ2/df < 3, RMSEA < 0.08, CFI > 0.9)
Human–Bird Interaction → REP	0.429 ***	0.362 ***	15.6%	Moderate Difference (10–20%)
REP → Mental Health	0.646 ***	0.689 ***	6.7%	Minor Difference (5–10%)
Direct Effect Intensity	0.323 ***	0.338 ***	4.6%	Minimal Difference (0–5%)
Indirect Effect Intensity	0.277 ***	0.25 ***	9.7%	Minor Difference (5–10%)
Total Effect Intensity	0.6 ***	0.588 ***	2%	Minimal Difference (0–5%)

*** means p < 0.001; Degree of Difference = |Total Sample Value − Online Sample Value|/Total Sample Value × 100%.

. We compared key indicators across four sample groups (total sample, online-only, offline-only, and weighted merged sample), covering descriptive statistics (Table A4), Pearson correlations (Table A5), mediation effects (Table A6), CFA reliability and validity (Table A7), and SEM core indicators (Table A8). The results confirmed high cross-sample consistency: the core variables’ descriptive statistics showed minimal to minor differences; key variable pair correlations were consistently significant with differences within 0–20%; mediation pathway coefficients and effect intensities maintained consistent directions without substantive deviations; CFA indicators including factor loadings Cronbach’s α AVE and CR exhibited differences ≤ 4.13%, with all constructs meeting reliability and validity standards; SEM fit indices of total and online-only samples all met academic criteria (χ²/df < 3, RMSEA < 0.08, CFI > 0.9) with consistent core pathway significance.

Notably, the small offline sample size precludes stable model construction in AMOS 26.0, so only Table A4, Table A5, Table A6, Table A7 and Table A8 include offline sample comparative analyses. These results, however, remain sufficient to demonstrate the offline sample’s consistency with the other samples and the core mediating mechanism’s robustness regardless of sample combination.

These multi-dimensional comparative analyses and sensitivity analyses complement the initial t-test and fully demonstrate that online and offline samples share consistent core characteristics and mechanisms. Merging the data does not introduce substantial bias, so the final valid questionnaire sample size is 321.

3.4. Analytical Methods

This study adopted a mixed approach strategy, integrating the strengths of latent variables modeling and the path analysis of manifest variables. We followed the recommendations of Wen, et al. [71], using AMOS 26.0 to validate the overall theoretical framework to ensure its validity and using PROCESS v4.1 to reveal the hierarchical mechanisms, with the results complementing rather than contradicting each other. AMOS 26.0 excels in latent variable modeling, verifying the integrity of the theoretical model through confirmatory factor analysis and overall model fit tests, though it is less flexible for analyzing differential mechanisms across interaction levels. PROCESS v4.1 specializes in manifest variable regression, enabling precise exploration of multi-group and moderated mediation to reveal fine-grained mechanisms, yet it cannot test overall model fit. The slight differences in mediation effect values, which result from methodological distinctions, do not contradict each other but jointly enhance the study’s rigor. The analysis process is as follows:

Firstly, SPSS 27.0 is used to preprocess the raw data, including missing value identification, outlier detection, and reverse scoring of reverse questions. Descriptive statistics were then performed, and the reliability of the scales (Cronbach’s α) was assessed to ensure the reliability of the measurement instrument.

Secondly, confirmatory factor analysis (CFA) and structural equation modeling (SEM) were conducted in AMOS 26.0 to evaluate the validity of the scale, test the fit of the overall theoretical model, and explore the mediating effect of the path of “human–bird interaction → restorative environmental perception (REP) → mental health”. In this model, human–bird interaction is regarded as a latent variable composed of sensory interaction, cognitive interaction, and participatory interaction. Interaction intensity (interaction frequency, duration, and bird–watching years) is introduced as an antecedent variable to investigate its predictive effect on human–bird interaction.

Finally, to further analyze the differential mechanism of different interaction levels, PROCESS v4.1 was used to construct multi-groups mediation models and moderated mediation models, with sensory interaction, cognitive interaction, and participatory interaction as independent manifest variables. We examined the mediating effect of REP between human–bird interaction and mental health, as well as the moderating effect of environmental characteristics on the path of “human–bird interaction → REP” under different levels of human–bird interaction to comprehensively reveal the mechanisms of human–bird interaction affecting mental health in urban wetland parks.

Maximum likelihood estimation (MLE) was used for all models. Bias-corrected Bootstrap methods (5000 repeated samples) were used for both mediation and moderation effects to calculate indirect effects and confidence intervals. The significance test criterion was α = 0.05, with a 95% confidence interval excluding 0 indicating statistical significance.

4. Results

4.1. Characteristics of the Data

This study collected a total of 321 valid questionnaires; the basic demographic information of the sample is presented in

Table A3. Questionnaire sample characteristics.

Content	Category	Frequency	Percentage	Content	Category	Frequency	Percentage
Gender	Male	148	46.1%	Bird–watching experience	Zero or occasional experience	147	45.8%
	Female	173	53.9%		Less than six months	28	8.7%
Age (years)	Under 18	46	14.3%		Six months to one year	36	11.2%
	18–29	187	58.3%		One to three years	56	17.5%
	30–39	55	17.1%		Over three years	54	16.8%
	40–59	24	7.5%	Frequency of interaction with birds	First time	53	16.5%
	Over 60	9	2.8%		Rarely (once a year)	99	30.9%
Educational Background	Junior high school and below	10	3.1%		Occasionally (once a quarter)	123	38.3%
	High school or vocational school	34	10.6%		Frequently (once a month)	37	11.5%
	Bachelor’s degree or associate degree	188	58.6%		Very frequently (once a week or more)	9	2.8%
	Graduate degree and above	89	27.7%
Identity (multiple choice)	Tourists	181	56.4%	Duration of interaction with birds	15–30 min	37	11.5%
	Bird–watchers	128	39.9%		30–60 min	68	21.2%
	Relevant scholars	35	10.9%		1–2 h	100	31.1%
	Volunteers	34	10.6%		2–3 h	58	18.1%
	Nature education practitioners	12	3.7%		More than 3 h	58	18.1%

. The gender distribution is relatively balanced, with males accounting for 46.1% and females for 53.9%. The age structure is dominated by the young adult population, with 58.3% aged 18–29 and 17.1% aged 30–39. The overall age range covers a wide range, meeting the typical visitor group structure of urban wetland parks. In terms of educational background, bachelor’s degree or associate degree accounted for the highest proportion (58.6%), graduate and above education accounted for 27.7%. The sample had high cultural quality, which was conducive to understanding and completing the questionnaire on human–bird interaction levels. In terms of identity, tourists (56.4%) and bird–watchers (39.9%) were the main ones, and there was a certain proportion of relevant scholars, volunteers, and nature education practitioners, reflecting the diverse motivations of human–bird interactions among respondents.

In addition, in terms of bird–watching experience, about 45.8% of the respondents had zero or occasional experience, 19.9% had some foundation but less than one year, and 34.3% had experience for over one year. This indicates that the sample included both a wide range of junior participants and a core population with continuous interactive experience. Most respondents interacted with birds less frequently (69.2% annually or quarterly), but the duration of interaction with birds was generally sufficient, with 67.3% lasting over one hour and 18.1% involving deep engagement lasting over three hours. In general, the sample structure is reasonable in terms of gender, age, and educational background, with a wide distribution of bird–watching experience levels. This ensures both differentiated measurement of interaction levels and the representativeness of regular visitors to urban wetland parks.

4.2. Reliability and Validity Analyses

To assess the reliability of the scale, this study uses SPSS 27.0 to calculate Cronbach’s α for each scale. Next, CFA was conducted on the overall measurement model via AMOS 26.0. In the mental health measurement scale, negative emotions had factor loads below 0.5 (0.07) and were therefore excluded from the overall measurement model. CFA was performed on the revised model, and the results showed acceptable model fit: χ²/df = 4.218, SRMR = 0.076, GFI = 0.903, AGFI = 0.842, CFI = 0.923. Cronbach’s α values for all variables exceeded 0.7 (

Table 1. Reliability and validity analyses of the scale.

Variable	Item	Factor Loading	Cronbach’s α	AVE	CR
Human–Bird Interaction	Sensory Interaction	0.855	0.917	0.712	0.879
	Cognitive Interaction	0.961
	Participatory Interaction	0.694
Environmental Characteristics	Visual Features	0.763	0.722	0.429	0.688
	Soundscape Features	0.653
	Facility Features	0.528
Restorative Environmental Perception	Being-Away	0.725	0.889	0.573	0.843
	Fascination	0.74
	Compatibility	0.812
	Extent	0.749
Mental Health	Positive Emotion	0.848	0.918	0.739	0.85
	Well-Being	0.871

) and the composite reliability (CR) was higher than 0.6 [72], indicating the good internal consistency of the scale. The standardized factor load values of most variables exceeded 0.7, and only a few items in facility features are slightly lower (the lowest is 0.528) but still within the acceptable range. The average variance extracted (AVE) for most latent variables is above 0.5, with environmental characteristics being slightly lower (AVE = 0.429) but still having convergent validity. As shown in

Table 2. Discriminant validity test of all constructs.

Variable	Human–Bird Interaction	Environmental Characteristics	Restorative Environmental Perception	Mental Health
Human–Bird Interaction	0.844
Environmental Characteristics	0.024	0.655
Restorative Environmental Perception	0.41	0.623	0.757
Mental Health	0.601	0.371	0.778	0.86

, the discriminant validity evaluation results show that, except for the high correlation between REP and mental health, the square root of AVE on the diagonal of other latent variables is greater than the corresponding correlation coefficient, indicating that most latent variables are interrelated but different from each other. Therefore, the discriminant validity of the scale as a whole is acceptable.

4.3. The Relationship Between Human–Bird Interaction and Mental Health and the Mediating Pathway

4.3.1. Overall Model Fitting and Path Analysis

An overall “interaction–perception–restoration” structural equation model was constructed via AMOS 26.0. The overall fitting of the model (Figure 4) was good: χ²/df = 2.938, RMSEA = 0.076, CFI = 0.957, GFI = 0.924, AGFI = 0.882. The results of AMOS 26.0 path analysis are shown in

Table 3. Estimated standardized coefficients.

Path	Estimate	C.R.
Human–Bird Interaction → Restorative Environmental Perception	0.429	6.757 ***
Human–Bird Interaction → Mental Health	0.323	6.314 ***
Restorative Environmental Perception → Mental Health	0.646	9.828 ***
Interaction Intensity → Human–Bird Interaction	1.022	8.503 ***

*** means p < 0.001.

.

• The predictive effect of interaction intensity on human–bird interaction

In order to verify whether the frequency, duration, and experience of individual contact with birds significantly affect the depth of human–bird interaction, this study takes interaction intensity as antecedent variable, and fits the path “interaction intensity → human–bird interaction”. The results of AMOS 26.0 path analysis (Table 3) show that interaction intensity has a significant positive predictive effect on human–bird interaction (β = 1.022, p < 0.001). This indicates that the more opportunities individuals have to interact with birds, the higher human–bird interaction level they achieve, validating the rationality of antecedent path. As shown in Figure 4, bird–watching experience (β = 0.84) and duration of interaction (β = 0.608) contributed more to interaction intensity.

2.

Relationships between human–bird interaction and restorative environment perception and mental health

In the main effects model (Figure 4), human–bird interaction had a significant positive effect on REP (β = 0.429, p < 0.001) and a direct positive effect on mental health (β = 0.323, p < 0.001), which supported hypotheses H1 and H2a. Moreover, we found that the deeper the human–bird interaction, the stronger the restorative perception of the environment and the higher the mental health level, with cognitive interaction (β = 0.944) and sensory interaction (β = 0.86) having the greatest impact.

3.

Relationship between restorative environment perception and mental health

REP also had a significant positive effect on mental health (β = 0.646, p < 0.001), confirming hypothesis H2b. This suggests that REP is not only the result variable of human–bird interaction, but may also serve as a mediating variable promoting mental health; hence, it is worth further exploring the underlying mediating mechanisms. As shown in Figure 4, all four dimensions of REP contribute to the assessment, with the dominant order being compatibility (β = 0.805) > fascination (β = 0.761) > being-away (β = 0.729) > extent (β = 0.727).

4.3.2. Mediating Effect of Restorative Environment Perception

The Bootstrap method (repeated sampling 5000 times) was used to further test the path of “human–bird interaction → REP → mental health” via AMOS 26.0. The results of the analysis (Figure 4 and

Table 4. The mediating effect results.

Variable	Effect Type	Value	Boot CI Lower Limit	Boot CI Upper Limit	Percentage
Human–Bird Interaction	Total	0.6 ***	0.507	0.683
	Direct	0.323 ***	0.189	0.455	53.83%
	Indirect	0.277 ***	0.162	0.404	46.17%

*** means p < 0.001.

) show that the 95% confidence intervals do not include zero, indicating that the mediating effect is significant, so hypothesis H2 was confirmed. REP plays a partial mediating role between human–bird interaction and mental health. Among them, the direct effect (0.323) and indirect effect (0.277) account for 53.83% and 46.17% of the total effect (0.6), respectively.

4.3.3. The Mediating Effects at Different Levels of Human–Bird Interaction

Having established the significant mediating role of REP in the overall model, we further investigated whether this mechanism operated uniformly across the different levels of human–bird interactions and the differences in the effect intensities. In order to examine the mechanism and intensity of different interaction levels, three mediation models were constructed using PROCESS v4.1 Model 4 (Figure 5). These models used one of the human–bird interaction levels (sensory interaction, cognitive interaction, and participatory interaction, respectively) as the independent variable, REP as the mediating variable, and mental health as the dependent variable. The paired difference test was conducted using data based on the percentile Bootstrap method (repeated sampling 5000 times), and the results are as follows.

There are significant differences in the direct effects of different levels of human–bird interaction on mental health (

Table 5. Comparison of differences in direct effect intensity in Model 4.

Interaction Level	Direct Effect β	Percentage	[95% Boot CI]	Comparison Object	Δβ [95% Boot CI]
Sensory Interaction	0.316 ***	61.93%	[0.233, 0.399]	Cognitive Interaction	−0.031 ** [−0.033, −0.029]
Cognitive Interaction	0.347 ***	66.32%	[0.267, 0.427]	Participatory Interaction	0.135 ** [0.133, 0.137]
Participatory Interaction	0.212 ***	55.38%	[0.127, 0.296]	Sensory Interaction	−0.104 ** [−0.106, −0.102]

Δβ means the difference in effect value; *** means p < 0.001; ** means p < 0.05.

). Cognitive interaction has the strongest direct effect on mental health (β = 0.347), significantly higher than sensory interaction (Δβ = 0.031, p < 0.05) and participatory interaction (Δβ = 0.135, p < 0.05). The direct effect of sensory interaction (β = 0.316) was also significantly higher than participatory interaction (Δβ = 0.104, p < 0.05). These results support hypothesis H1a.

There were also statistical differences in the intensity of the mediating effect of REP across different levels of human–bird interaction (

Table 6. Comparison of differences in indirect effect intensity in Model 4.

Interaction Level	Indirect Effect β	Percentage	[95% Boot CI]	Comparison Object	Δβ [95% Boot CI]
Sensory Interaction	0.194 ***	38.07%	[0.129, 0.265]	Cognitive Interaction	0.018 ** [0.016, 0.019]
Cognitive Interaction	0.176 ***	33.68%	[0.107, 0.250]	Participatory Interaction	0.005 ** [0.004, 0.007]
Participatory Interaction	0.171 ***	44.62%	[0.098, 0.249]	Sensory Interaction	−0.022 ** [−0.024, −0.021]

Δβ means the difference in effect value; *** means p < 0.001; ** means p < 0.05.

). The indirect effect of sensory interaction on mental health through REP was the strongest (β = 0.194), significantly higher than cognitive interaction (Δβ = 0.018, p < 0.05) and participatory interaction (Δβ = 0.022, p < 0.05). Moreover, the indirect effect of cognitive interaction (β = 0.176) was significantly higher than participatory interaction (Δβ = 0.005, p < 0.05). These results support hypothesis H2c.

4.4. Moderating Effect of Environmental Characteristics

4.4.1. Overall Moderated Mediation Effect

To examine the moderating effect of environmental characteristics, a moderated mediation analysis (PROCESS v4.1 Model 7; Figure 6) was conducted with human–bird interaction as the independent variable, REP as the mediating variable, mental health as the dependent variable, and environmental characteristics as the moderating variable. The results showed that human–bird interaction had a significant positive effect on REP (β = 0.324, p < 0.001), and environmental characteristics directly and positively influenced REP (β = 0.51, p < 0.001). However, the interaction term between human–bird interaction and environmental characteristics did not significantly influence REP (β = −0.029, p = 0.483). These results suggest that environmental characteristics do not significantly modulate the pathway between human–bird interaction and REP. The index test further confirmed that the mediating effect was not significant (Index = −0.015, Boot CI [−0.047, 0.024]). Therefore, hypothesis H3 was not supported.

4.4.2. The Moderated Mediation Effects at Different Levels of Human–Bird Interaction

One of the human–bird interaction levels (sensory interaction, cognitive interaction, participatory interaction, respectively) was used as the independent variable, REP was used as the mediating variable, mental health was used as the dependent variable, and environmental characteristics was used as the moderating variable to conduct moderated mediation analysis (PROCESS v4.1 Model 7; Figure 6). The results showed that the interaction term between sensory interaction and environmental characteristics did not significantly influence REP (β = 0.007, p = 0.873), and the moderating index was not significant (Index = 0.0037, Boot CI [−0.032, 0.047]). The results of the moderated mediation analysis of cognitive interaction (β = −0.03, p = 0.472; Index = −0.016, Boot CI [−0.052, 0.025]) and participatory interaction (β = −0.029, p = 0.497; Index = −0.017, Boot CI [−0.063, 0.031]) were also insignificant. The mediating effects of three levels of interaction (human–bird interaction → REP → mental health) were all significantly positive (Boot CI did not include 0) but not moderated by environmental characteristics.

To further verify the robustness of the non-significant moderating effect, a stratified conditional process analysis was conducted by estimating the indirect effects of human–bird interaction on mental health through REP at high (+1 SD) and low (−1 SD) levels of environmental characteristics. As shown in

Table 7. Comparison of differences in indirect effect intensity in Model 7.

Independent Variable	β (Indirect Effect at +1 SD)	β (Indirect Effect at +1 SD)	Δβ [95% Boot CI]	Significance (Δβ)
Human–Bird Interaction	0.185 ***	0.155 ***	−0.03 [−0.093, 0.047]	Not Significant
Sensory Interaction	0.173 ***	0.166 ***	0.007 [−0.064, 0.094]	Not Significant
Cognitive Interaction	0.147 ***	0.179 ***	−0.032 [−104, 0.049]	Not Significant
Participatory Interaction	0.121 ***	0.155 ***	−0.034 [−127, 0.061]	Not Significant

Indirect effects were estimated using 5000 bootstrap samples (PROCESS v4.1 Model 7); *** means p < 0.05; Δβ = β(indirect effect at +1 SD) − β(indirect effect at −1 SD); ±1 SD indicates one standard deviation above or below the mean of the moderator (environmental characteristic).

, the indirect effects remained significant at both high and low environmental characteristics levels. However, the differences in effect size (Δβ) were small and all 95% confidence intervals included zero, indicating that environmental characteristics did not significantly moderate the mediation pathway.

Through multi-group mediation moderated mediation analyses, it was found that environmental characteristics did not significantly moderate the relationship between different human–bird interaction levels and REP. Although the indirect effects of some groups under low environmental characteristics were slightly higher than those under high environmental characteristics, the differences were not significant (Table 7). This indicated that the restorative benefits of human–bird interaction are relatively stable across different environmental conditions. Therefore, hypothesis H3a was not supported.

5. Discussion

5.1. Interaction–Perception–Restoration Mechanism

Based on the landscape perception theory, this study constructed a theoretical model of “interaction–perception–restoration” to explore the relationship among human–bird interactions, REP, environmental characteristics, and mental health in urban wetland parks. Following the model, we proposed eight hypotheses and tested them (

Table 8. Hypothesis testing results.

Hypothesis	Results
H1: Human–bird interactions have significant positive impacts on mental health.	Supported
H1a: The effect intensity of different levels of human–bird interactions on mental health differs.	Supported
H2: REP plays a mediating role between human–bird interactions and mental health.	Supported
H2a: Human–bird interactions have a significant positive effect on REP.	Supported
H2b: REP has a significant positive effect on mental health.	Supported
H2c: The mediating effect intensity of REP differs across the relationships between different levels of human–bird interactions and mental health.	Supported
H3: Environmental characteristics moderate the relationship between human–bird interactions and REP.	Rejected
H3a: The moderating effect of environmental characteristics differs across the relationships between different levels of human–bird interactions and REP.	Rejected

). The results showed that all other hypotheses were validated, except for the moderating effect hypothesis of environmental characteristics (H3, H3a), which revealed the complex relationships between human–bird interactions and mental health.

This study confirms that human–bird interactions have both direct and indirect effects on mental health, with 46.17% of the total effects mediated by REP, thereby validating the rationality of the theoretical framework of “interaction–perception–restoration”. This study expands the limitations of existing studies that merely treat birds as static environmental elements (bird presence [14,15], bird diversity and richness [19,20,21], bird sounds [23,24], etc.), revealing the profound psychological value of dynamic human–bird interactions. Human–bird interaction is not merely a simple form of natural exposure, but rather an active process of psychological restoration through cognitive processing and emotional activation (understanding birds [12], connecting with nature [51,52], etc.). This aligns with the perspective in environmental psychology that emphasizes individuals’ initiative in actively interacting with the environment [74].

The mediating role of REP highlights the central role of subjective experience in the relationship between human–bird interactions and mental health. Among its dimensions, compatibility (β = 0.805) and fascination (β = 0.761) contribute the most, indicating that human–bird interactions that match individual needs (such as interests and abilities) or stimulate active attention are the keys to the transformation of mental health benefits. This is consistent with the view proposed by Kaplan [75] that “the meaning-making of nature experiences is central to restoration” and also provides new evidence for the applicability of landscape perception theory in dynamic biological interaction scenarios.

5.2. Differentiated Mental Health Benefits of Different Levels of Human–Bird Interactions

The study found significant differences in mental health effects at different levels of interactions, which provides a new perspective for understanding the benefits of human–bird interactions. The direct effect of cognitive interaction was the strongest (β = 0.347), indicating that higher cognitive activities such as identifying birds and inferring their behaviors could enhance meaning-making and focus attention, directly promoting positive emotions and well-being. This aligns with the conclusion by Randler, et al. [76] that deep cognitive engagement can strengthen emotional connections between humans and birds, thereby contributing to enhancing well-being. Sensory interaction produced the largest indirect effect through REP (β = 0.194), supporting the assertion of ART that unconscious sensory stimuli (e.g., bird sounds, plumage colors) are more likely to trigger “non-compulsive attention” [11], promoting the rapid generation of “being-away” and “fascination”.

In contrast, the overall effect of participatory interaction is relatively weak, a finding that can be theoretically explained by two environmental psychology mechanisms closely related to cognitive load and benefit types. First, as per ART, psychological restoration is optimized under low cognitive loads [11]. Participatory activities (e.g., bird rescue, citizen science surveys) require goal-directed effort (e.g., following rescue protocols, recording data) that imposes short-term cognitive strain. This strain partially offsets immediate restoration gains, aligning with the conclusion by Barton and Pretty [77] that short-duration, lower-intensity natural engagement yields more pronounced immediate mood benefits, while higher-intensity or more effortful involvement leads to diminishing short-term returns. Second, environmental psychology distinguishes between short-term affective benefits and long-term eudaimonic benefits [78]. This dual-benefit framework explains why participatory interaction’s weak short-term effect but does not negate its value. Participatory interaction (e.g., bird rescue, citizen science) may generate deeper and more enduring mental health benefits (e.g., sense of life meaning, self-efficacy) by fostering environmental responsibility, belonging, and social connection [51,52], which requires further validation through longitudinal studies in the future.

5.3. The Non-Significant Moderating Effect of Environmental Characteristics and the Central Role of Human–Bird Interactions

The rejection of H3 and H3a, which proposed the moderating role of environmental characteristics, may result from several interconnected factors. First, the average variance extracted (AVE) of environmental characteristics was 0.429, slightly below the ideal threshold of 0.5, indicating a potential limitation in measurement validity that might have weakened the statistical sensitivity of the moderating pathway. This limitation may be exacerbated by the second factor: the study site. Haizhu National Wetland Park is a well-managed and ecologically high-quality urban wetland with relatively homogeneous environmental conditions, such as a consistent landscape structure and well-maintained facilities [21]. This homogeneity reduced the variability required for testing moderation. In addition, previous studies have shown that when environmental quality exceeds a certain threshold, its incremental contribution to psychological restoration tends to level off [77,79,80]. This does not contradict the significant direct effect of environmental characteristics on restorative environmental perception (β = 0.51) observed in this study. Rather, it suggests that while good environmental quality remains a necessary foundation for restoration, its ability to further strengthen the relationship between human–bird interactions and restorative perception becomes limited once a high baseline quality has been achieved. Under such circumstances, the cognitive and emotional engagement stimulated by human–bird interactions may play a more decisive role in generating restorative experiences than additional environmental optimization [81,82].

Future research could further examine the contextual boundaries of this moderating effect by incorporating other potential moderators or subgroups. Variables such as individual nature connectedness, bird–watching frequency and expertise, or wetland parks with different management intensities and environmental gradients may provide greater heterogeneity for identifying moderation effects. Multi-site or longitudinal studies could also help determine the conditions under which environmental attributes amplify or constrain restorative mechanisms.

Overall, although environmental characteristics significantly and directly enhance restorative environmental perception, their moderating role appears limited in high-quality environments where ecological and design baselines are already well established. This finding refines the understanding of environmental quality as a necessary but not decisive factor. The key driver of mental health restoration lies in the quality and depth of human–bird interactions themselves, in which cognitive processing and emotional activation (such as concentration, curiosity, and pleasure) play a central role [28,43,45]. This conclusion is consistent with previous environmental psychology studies, which emphasize that meaningful and diverse interactions with nature contribute more strongly to attention restoration and emotional recovery than incremental physical improvements in already excellent environments [77,79,80,81].

5.4. Recommendations for the Design and Service Management of Human–Bird Interactions in Wetland Parks

In response to the differentiated effects of different interaction levels, urban wetland parks can develop hierarchical strategies: (1) For sensory interaction: Set up 360-degree bird–watching platforms at key locations to enhance visual accessibility; install noise reduction barriers along main trails to preserve natural soundscapes; add signposts guiding visitors to adjust postures for optimal observation. (2) For cognitive interaction: Offer telescope rental services and install augmented reality bird identification devices that deliver real-time species information. In addition, compile and distribute localized bird guidebooks and organize regular expert-led lectures to strengthen public understanding of bird behavior and ecology. (3) For participatory interaction: Initiate a “Bird Habitat Maintenance Volunteer Program” with monthly activities; develop a citizen-science data platform that recognizes active contributors through equipment rewards; and design a tiered nature education curriculum for different age groups to promote progressive learning and engagement. Emphasizing feedback on conservation outcomes can further enhance participants’ long-term motivation, balancing short-term management inputs with sustained social and ecological benefits.

In addition, based on the mediating role of restorative perception, it is necessary to focus on improving “compatibility” and “fascination”. To strengthen “compatibility”, spatial organization should better match visitors’ abilities, preferences, and behavioral rhythms. This can be achieved by establishing differentiated bird–watching trails with varying lengths, rest points, shaded shelters, and barrier-free routes for children, the elderly, and people with limited mobility. Seating areas oriented toward major bird habitats can provide opportunities for quiet observation, while route segmentation and loop design can balance accessibility with habitat protection. To enhance “fascination”, create multi-sensory and story-based experiences that deepen attention and emotional resonance. Examples include installing seasonal interpretive boards highlighting current migratory species, designing interactive installations that reproduce bird calls, and curating thematic events such as “wetland bird weeks” that combine observation, drawing, and short guided walks. These approaches stimulate curiosity and attentional focus, reinforcing the affective pathways of restoration.

Furthermore, considering the moderating role boundary of environmental characteristics, management in high-quality wetlands should prioritize interaction-oriented improvement over marginal physical environment enhancements. Instead of expanding built infrastructure, emphasis should be placed on improving the quality, accessibility, and continuity of interaction opportunities. For instance, lowering viewing heights, ensuring clear sightlines to bird feeding areas, and designing vegetated buffer zones that allow proximity without disturbance. At the same time, maintaining a stable baseline of environmental quality through regular monitoring, vegetation management, and noise control remains essential to ensure that interaction-focused initiatives operate within a healthy ecological framework.

These recommendations emphasize the dual ecological and recreational roles of urban wetland parks, integrating biodiversity conservation with nature-based recreation through improved interactive experiences.

5.5. Limitations

This study has three limitations. Firstly, our research site was restricted to the Haizhu National Wetland Park in Guangzhou, which may limit generalizability to wetland parks in other climatic zones or with different management levels. The sample is mainly composed of young people (58.3% aged 18–29) and highly educated people (86.3% with a bachelor’s degree or above). These characteristics are partly due to Haizhu Wetland’s strong science popularization orientation, which attracts highly educated groups and results in higher participation willingness among young enthusiasts in online bird–watching communities. The experience of middle-aged people, elderly people, low-education groups, and special groups (such as disabled people) are not fully covered. Meanwhile, the data are dominated by online responses, with only 39 valid field survey samples, leading to potential bias from single data source. Future research should expand to multi-site investigations across different climatic zones and management levels and enrich the sample structure to include diverse populations (e.g., older adults, lower education groups, disabled individuals) for improved generalizability. Additionally, future studies should either have increased offline sample size or conduct separate analyses of online and offline samples to improve generalizability and capture scenario-specific mechanisms.

Secondly, the cross-sectional design of the study cannot reveal long-term dynamics of the mental health benefits of human–bird interactions (e.g., the impacts of seasonal variations on interaction quality, cumulative mental health benefits of long-term participation), nor clarify differences in the moderating role of environmental characteristics across different time scales. In the future, longitudinal empirical research needs to be expanded.

Thirdly, the key variables of this study are mainly self-reported perception indicators. The lack of objective environmental indicators may fail to capture the real environmental constraints. Additionally, the current exploration of environmental characteristics focuses on the overall level without subdividing subdimensions such as visual landscapes, auditory environments, and facility conditions. This overlooks potential differentiated moderating effects of subdimensions on the core pathway. Subsequent studies could integrate physiological indicator measurements, objective environmental data collection, interaction behavior recording, and third-party evaluations to further enhance the objectivity of the research and the precision of environmental mechanism analysis, while exploring differentiated effects of environmental characteristic subdimensions.

6. Conclusions

Based on the “interaction–perception–restoration” theoretical framework, this study systematically explored the impact mechanism of human–bird interactions on mental health in urban wetland parks. The main conclusions are as follows:

(1)

Human–bird interactions have significant positive effects on mental health, and there are two paths of “direct effect” and “indirect effect”. Among them, the mediating effect of restorative environmental perception accounts for 46.17% of the total effect.

(2)

There are differences in the effects of different human–bird interaction levels: cognitive interaction exhibits the strongest direct effect, sensory interaction produces the largest indirect effect through restorative environmental perception, and participatory interaction has relatively weaker positive effects but may hold long-term potential.

(3)

Environmental characteristics have direct positive effects on restorative environmental perception but do not play a significant moderating role between human–bird interactions and restorative environmental perception. This indicates that, in the context of relatively high-quality natural environments such as urban wetland parks, human–bird interactions themselves are the core driver of mental health restoration. This does not diminish the importance of environmental quality, but clarifies its role boundaries.

This study refined the hierarchical structure of human–bird interactions, verified the mediation mechanism of restorative environmental perception, and clarified the role boundary of environmental characteristics. We provided a new perspective for understanding the relationship between dynamic human–bird interactions and mental health, as well as a practical basis for ecological service optimization and public health promotion of urban wetland parks, helping to jointly improve biodiversity conservation and human well-being.