Designing Smart Urban Parks with Sensor-Integrated Landscapes to Enhance Mental Health in City Environments

Abstract

As mental health issues such as stress, anxiety, and depression become increasingly prevalent in urban populations, there is a critical need to embed restorative functions into the built environment. Urban parks, as integral components of ecological infrastructure, play a vital role in promoting psychological well-being. This study explores how diverse park environments facilitate mental health recovery through multi-sensory engagement, using integrated psychophysiological assessments in a wetland park in Zhengzhou, China. Electroencephalography (EEG) and perceived restoration scores were employed to evaluate recovery outcomes across four environmental types: waterfront, wetland, forest, and plaza. Key perceptual factors—including landscape design, spatial configuration, biodiversity, and facility quality—were validated and analyzed for their roles in shaping restorative experiences. Results reveal significant variation in recovery effectiveness across environments. Waterfront areas elicited the strongest physiological responses, while plazas demonstrated lower restorative benefits. Two recovery pathways were identified: a direct, sensory-driven process and a cognitively mediated route. Biodiversity promoted physiological restoration only when mediated by perceived restorative qualities, whereas landscape and spatial attributes produced more immediate effects. Facilities supported psychological recovery mainly through cognitive appraisal. The study proposes a smart park framework that incorporates environmental sensors, adaptive lighting, real-time biofeedback systems, and interactive interfaces to enhance user engagement and monitor well-being. These technologies enable urban parks to function as intelligent, health-supportive infrastructures within the broader built environment. The findings offer evidence-based guidance for designing responsive green spaces that contribute to mental resilience, aligning with the goals of smart city development and healthy life-building environments.

1. Introduction

The 21st century is witnessing significant global health challenges, characterized by rising physical inactivity, non-communicable diseases, and adverse physiological and psychological symptoms. The COVID-19 pandemic exacerbated this concern, with an additional 53 million cases of major depressive disorder reported worldwide in 2020 [1]. Consequently, urban populations increasingly face pervasive mental health burdens, with chronic stress, anxiety, and depression are now recognized as critical public health priorities. Recent epidemiological studies reveal that over 40% of city dwellers exhibit clinically significant mood disturbances, driven by factors such as high-density living, digital overload, and environmental stressors [2]. Among urban residents, young adults represent a particularly vulnerable yet often overlooked group. While navigating a unique transitional period marked by psychological socialization challenges [3], they are not typically categorized alongside children or the elderly as vulnerable populations for physical and mental health conditions. This perception gap resulted in insufficient research focus and support for their specific needs. In response to these escalating challenges, global bodies such as the World Health Organization (WHO) and the United Nations Human Settlements Programme (UN-Habitat) emphasized the urgent need to integrate health-promoting technologies into urban infrastructure, highlighting the critical role of smart built environments in supporting psychological resilience [3]. This imperative aligns closely with the growing recognition of nature-based solutions (NbS) as essential strategies for addressing interconnected urban crises, including public health and climate adaptation. This alignment is especially pertinent in the post-pandemic era, where accessible, restorative environments have proven vital for community well-being and resilience [4].

Given the growing public health challenges, recent research has begun to explore the potential of urban nature as a cost-effective means of addressing these issues. Urban parks, encompassing general vegetation levels as well as managed green and blue spaces, have been shown to significantly improve public physical and mental health by reducing stress, enhancing life satisfaction, and fostering social interaction [5,6]. Complementary studies further establish that green space exposure mitigates negative affective states, including aggression, frustration, and anger [7]. Critically, when integrated with real-time biometric monitoring and adaptive systems, these spaces evolve beyond passive greenery into active health interventions [5]. Empirical evidence confirms that a technologically augmented nature reduces cortisol levels, optimizes mood regulation, and accelerates cognitive restoration [3]. Among urban green spaces, wetland parks represent particularly promising therapeutic landscapes due to their multisensory richness [8]. The dynamic interplay of aquatic ecosystems may generate distinctive restorative stimuli, potentially exceeding those typically found in conventional park settings [9]. Despite growing recognition of urban parks’ therapeutic value, two critical research limitations persist: first, existing studies predominantly rely on simulated environments, while operational parks—especially complex wetland systems—lack embedded sensing capabilities to quantify user engagement and psychophysiological responses (e.g., physiological measurements and psychological assessments) across heterogeneous zones; second, the mechanisms through which different perceptual dimensions, including landscape aesthetics, spatial organization, and biodiversity appreciation, interact to influence well-being outcomes remain poorly understood. These knowledge gaps fundamentally impede the development of evidence-based design strategies leveraging multi-sensory biofeedback data for optimization.

To better integrate sensor-integrated landscapes into smart park design, it is essential to deeply investigate the underlying mechanisms by which multi-dimensional perception influences physiological and psychological restoration. This study selected Zhengzhou City as the research area. As the capital city of Henan Province, Zhengzhou experienced rapid urbanization in recent years, characterized by high population density, extensive transportation networks, and rising daily life and work pressures [10]. Within such a fast-paced urban context, young adults face growing psychological burdens, with many suffering from anxiety, depressive symptoms, or physical discomfort due to intense academic pressures. Therefore, examining how urban parks can alleviate psychological and physiological stress among young adults holds significant practical relevance. Furthermore, Zhengzhou, as a typical plain city with a dense network of rivers and waterways, boasts numerous parks and green spaces along water bodies, including notable wetland ecosystems such as Longhu and Xianghu [11,12,13]. The ecological beauty surrounding these waterways attracts many young adults seeking weekend outings, excursions, and recreational activities. In this study, we explored the differences in the physiological (EEG-α/β ratio) and psychological (mental restoration scores) restorative outcomes among young adults exposed to four urban wetland park environments. Subsequently, we delved into the impact pathways of multi-dimensional perception (i.e., landscape perception, spatial perception, facility perception, and biodiversity perception) on young adults’ EEG-α/β ratio and mental restoration scores to clarify the mechanism of a wetland park in triggering physiological and psychological restoration. The aims of this study are as follows:

• To quantify restorative outcomes linked to perceptual factors (landscape design, spatial configuration, biodiversity, and facilities);
• To identify distinct psychophysiological pathways (direct sensory vs. cognitively mediated) for mental health recovery;
• Thereby, to propose a smart park framework leveraging environmental sensors, adaptive systems, and biofeedback.

2. Theoretical Framework

2.1. Restorative Environmental Theory

Urban natural environments have been widely recognized for their restorative effects on human well-being [14]. Key theories explain these effects: attention restoration theory (ART) [15], stress reduction theory (SRT) [16], and prospect and refuge theory (PRT) [17].

ART posits that nature alleviates mental fatigue by providing “soft fascinations” (effortlessly engaging stimuli), a sense of “being away,” and fostering nature connectedness, thereby restoring attention [18,19]. Routine tasks often demand prolonged effort and focused attention, leading to mental exhaustion and impaired cognitive function [15]. In contrast, nature-rich environments provide a restorative effect by offering soft fascinations—stimuli that capture attention effortlessly. These environments also create a sense of being away from daily life and foster a connection with nature, thereby reducing mental fatigue and enhancing cognitive performance [19]. Complementing ART, SRT emphasizes nature’s role in reducing physiological arousal, negative affect, and stress-induced impairments, while promoting positive emotions [16]. PRT explains environmental preferences, suggesting that people are drawn to settings offering both prospect (open views) and refuge (secure shelter), finding balanced prospect–refuge environments most attractive and satisfying [17,20].

Attention restoration, stress reduction, and psychological comfort are integral to restoration. Although emphasizing different aspects, these theories overlap and reinforce each other [21]. Natural settings providing prospect (open vistas) and refuge (secure spots) contribute to restoration. PRT’s spatial configuration supports ART by enabling “extent” (coherent space) and “being away.” SRT’s stress reduction is amplified in PRT-designed environments where prospect–refuge balance reduces threats and enhances safety [22]. These interactions represent the holistic nature of a restorative environment [23], addressing cognitive fatigue, stress response, and spatial assessment.

Collectively, these theories demonstrate that exposure to well-designed nature enhances emotional comfort, attentional restoration, and stress reduction. However, how specific natural settings engage these mechanisms remains inadequately understood. Wetland environments, in particular, offer unique sensory and ecological characteristics—water movement sounds, biodiverse visual complexity, and transitional edge habitats—that may simultaneously activate multiple restorative pathways [24]. Yet, empirical studies on wetland parks’ restorative efficacy remain limited. Current research predominantly focused on forests or general urban green spaces when investigating restorative mechanisms in natural environments [25]. Unlike these relatively homogeneous settings, wetlands encompass diverse subtypes (e.g., riparian, marsh, and constructed ponds) with distinct ecological attributes. This heterogeneity challenges generalized conclusions, as restorative pathways may vary significantly across aquatic-dominated environments [26].

2.2. Perception, Preference, and Restoration

Perceptual processes fundamentally mediate environmental restoration, as physical characteristics often diverge from perceived quality [27]. Consequently, perceptual metrics are essential for capturing dimensions beyond biophysical measurements. In urban contexts, various frameworks evaluate park characteristics and human preferences [28]. For instance, Kaplan and colleagues established a preference matrix based on human needs and information availability, using four key informational variables to assess features and predict perceptions. Additionally, Grahn and Stigsdotter’s (2010) “Perceived Sensory Dimensions” (PSD) evaluate urban green space qualities across eight dimensions [29]. Furthermore, Mikel Subiza-Perez proposed the “Perceived Environmental Aesthetic Qualities Scale” for evaluating green-blue spaces [30]. Moreover, physical attributes critically shape these perceptions. Empirical studies consistently demonstrate that vegetation density, water visibility, and amenity provision significantly influence landscape evaluations [31,32,33].

Preference, reflecting the desire for certain landscapes, is the product of perception and is often assessed through attractiveness, aesthetic quality, or scenic beauty [34]. Viewing pleasing natural scenery can block negative thoughts and achieve psychological restoration. Kaplan and Kaplan (1989) further highlight that pleasant environments allow individuals to relax with effortless attention and restore directed attention capacity, alleviating mental fatigue [15]. Each urban park possesses unique ecological and landscape characteristics that contribute to its meaning and identity. Individuals perceive, experience, and value these spaces based on their personal perceptions and associated emotions [35]. When space is recognized as meaningful and identifiable, it fosters intangible, emotional, and cognitive bonds that positively impact mental restoration and well-being [36].

Despite established perceptual frameworks, current research exhibits a tendency to evaluate landscape dimensions. Studies predominantly isolate singular variables (e.g., visual greenery or spatial configuration), neglecting synergistic interactions between sensory, ecological, and aesthetic domains [37]. This gap is most pronounced concerning biodiversity perception—a multidimensional construct requiring concurrent assessment of species richness discernibility (visual), biophonic complexity (auditory), and habitat heterogeneity (cognitive). Although biodiversity perception gained attention, it is rarely integrated within a multi-dimensional perceptual framework. Therefore, research on multi-dimensional perception is needed to inform evidence-based optimization of multifunctional landscape design.

2.3. Assessment of Physiological and Psychological Restoration

To empirically validate restorative mechanisms proposed by ART, SRT, and PRT, multidimensional assessment approaches are essential. Physiological recovery targets autonomic nervous system regulation, which is quantified using biomarkers [38]. Common tools include blood pressure monitors (tracking stress responses), EDA devices (measuring skin conductance for arousal assessment), and EEG (recording brain activity) [39]. Recent EEG advancements enable real-world applications; an Edinburgh field study used portable EEG to monitor neural indicators across urban environments [40,41]. Results show that green spaces reduced frustration/arousal while enhancing meditation, validating the EEG’s efficacy in quantifying physiological restoration mechanisms. Psychological restoration is commonly assessed using self-report questionnaires [42]. These tools provide valuable insights into individuals’ subjective perceptions and experiences. Some widely used questionnaires include the Perceived Stress Scale (PSS), which measures perceived stress levels and coping abilities; the Positive and Negative Affect Schedule (PANAS), which evaluates positive and negative emotional states; and the Profile of Mood States (POMS), which assesses transient mood states [43]. These tools have been validated across diverse populations and settings.

The Perceived Restorativeness Scale (PRS), developed by Hartig and refined by Pasini et al., serves as a pivotal metric integrating ART, SRT, and PRT [44]. This tool can quantify the interplay between environmental settings, human perception, and restorative outcomes. The PRS supports ART by assessing how much an environment can draw in and hold a person’s attention while also promoting mental rejuvenation and cognitive restoration [22]. Similarly, the PRS captures the qualitative elements that promote relaxation, emotional balance, and stress reduction by evaluating coherence, compatibility, and the sense of being away, which are aspects of PRS, closely fitting with the principles of SRT [45]. The scale incorporates the ideas of PRT by taking into consideration how particular contextual factors, such as coherence and compatibility, contribute to a feeling of security and emotional comfort [46]. Empirical studies confirm that PRS mediates environmental features with physiological and psychological restoration [47,48,49], enabling evidence-based design for aesthetic, cognitive, and affective optimization.

While previous studies examined the combination of physiological measurements and self-reported questionnaires for evaluating recovery outcomes, several research gaps persist. First, the influence of specific green space attributes on physiological restoration remains underexplored, with limited evidence on how distinct environment types modulate health recovery processes. Existing physiological studies predominantly compare generic “natural versus urban” environments rather than analyzing nuanced typologies. Second, the differential pathways through which urban parks facilitate physiological versus psychological restoration remain poorly characterized. These limitations highlight the critical need for multimodal approaches integrating EEG, behavioral tracking, and psychometric assessments to comprehensively evaluate restoration mechanisms.

2.4. The Current Study

To bridge the previously mentioned knowledge gaps, we establish a hierarchical research framework progressing from empirical validation to smart design translation (Figure 1). First, linear regression validates whether objective landscape elements underpin young adults’ perceptual evaluations (spatial/aesthetic/biodiversity/facility dimensions). Second, multimodal assessment (EEG biomarkers + Perceived Restoration Scales) quantifies restoration differentials across four designed environments (waterfront/wetland/forest/plaza). Third, structural equation modeling deciphers dual mediation pathways. Finally, these mechanistic insights inform an IoT-driven smart park framework, deploying environmental sensors, adaptive systems, and biofeedback interfaces to optimize perceptual experiences and amplify mental health benefits—translating evidence into precision design.

3. Methods

3.1. Study Area and Study Sites

Zhengzhou, the capital of Henan Province, spans the Yellow River and Huai River basins and is characterized by a plain–river network, with rivers, canals, and artificial lakes interweaving to structure the city’s open-space network. The climate is a monsoon-influenced humid subtropical climate, with hot, humid summers and cool, dry winters. With a population exceeding ten million and a high level of urbanization, the city’s blue–green infrastructure undertakes ecological regulation and water quality improvement while accommodating intensive daily recreation and commuting. Against this backdrop, most urban parks in Zhengzhou are aligned along river–lake corridors, forming a continuous “river–lake–greenway” open-space network [11]. Along the dimensions of hydro-geomorphology/urban water systems and the population structure, Zhengzhou is highly similar to many plain–river network/delta/lowland lakeside metropolitan areas worldwide.

Beilonghu Wetland Park, located in Jinshui District, was selected as the study area (Figure 2). It covers approximately 164,000 m², including about 38,800 m² of water bodies and 86,500 m² of green space. Developed through regional water system remediation and ecological restoration, the park is among the largest constructed wetland parks in the city; spatially, it comprises a mosaic of water, wetland, grassland, and woodland patches, incorporating all key environmental attributes typical of wetlands. Integrating flood regulation, primary purification, aquatic ecosystem restoration, and public recreation, it is well-suited for investigating restorative responses along gradients of naturalness, openness, and water cues. Accordingly, the perceptual preference-driven–driven dual-pathway restoration mechanism shows good cross-city transferability and reusability under appropriate contextual conditions.

To ensure that the experimental sites captured representative differences in environmental types, landscape features, biodiversity levels, and the spatial arrangement of vegetation and water, six experts conducted on-site reconnaissance and assessment. Based on site representativeness, four experimental zones were selected to represent core ecological and functional gradients within urban wetland parks: wetland, plaza, waterfront, and forest (Figure 3). This stratification captures the following: (1) a hydrological continuum from aquatic-dominated to terrestrial systems; and (2) a structural complexity spectrum ranging from open vistas to densely vegetated habitats. Fieldwork was conducted over 15 consecutive days in October 2024, with two daily sessions (09:00–13:00 and 15:00–17:00).

3.2. Environmental Data Collection

To quantify the visual composition of the surrounding environment at each study site, a series of 180° horizontal panoramic photographs was captured. Photographs were taken using a digital camera mounted on a tripod at a height of approximately 1.6 m at the designated sampling point within each scene. Several common landscape elements, namely architecture, sky, vegetation, water, roads, facilities, and landscape richness, were used for statistical analysis. The first six features refer to the proportion of a certain landscape element in the image, while the last feature refers to their mixture. The formula for calculating richness uses the Shannon entropy index, which is as follows:

$$richness=-\sum p_i\ast ln\left(p_i\right)/ln\left(k\right)$$

where pi is the proportion of element i, and k is the number of elements. Plant species diversity was quantified within fixed-area plots centered at each sampling location (radius = 50 m). All vascular plant species were identified and recorded. Species coverage (%) was estimated using a modified Braun-Blanquet scale, and individual counts per species were derived from coverage area conversion. Avian diversity was assessed via standardized point-count surveys. Two professional ornithologists independently recorded bird species and individual counts within a 50 m radius of each sampling location during participant observation sessions. Biodiversity was quantified using the Shannon Diversity Index (H’), which is as follows:

$$H\prime =-{\displaystyle \sum }_{i=1}^S{p}_{i}ln\left(p_i\right)$$

where S is the total number of species, p_i is the proportion of individuals belonging to species i, and ln denotes the natural logarithm. Higher H′ values indicate greater species diversity.

3.3. Psychophysiological Assessment

3.3.1. Preparation of Participants

Sample size was determined a priori using G*Power 3.1 (effect size *f* = 0.25, power = 0.95, and α = 0.05) [50], indicating ≥50 participants required. Therefore, we recruited 68 young adults, and the inclusion criteria were as follows: participants aged between 18 and 25 years, right-handed, and available to attend all scheduled experimental sessions. The exclusion criteria included severe visual impairments or clinical diagnoses of psychiatric, neurological, or cognitive disorders. After exclusions, 59 valid participants completed all sessions. Among the valid participants, there were 35 males (57.63%) and 24 females (42.37%), with ages ranging from 18 to 24 years and above. Specifically, 9 participants (15%) were aged between 18 and 20 years, 40 participants (68%) between 21 and 23 years, and 10 participants (17%) were aged 24 years or older. All experiments adhered strictly to relevant ethical guidelines and regulations and were conducted only after obtaining informed consent from all participants.

3.3.2. Measurement of Physiological Indicators

EEG data were collected using a non-invasive Emotiv EPOC + headset (AEMOTIV Inc., San Francisco, CA, USA), with recordings wirelessly transmitted to a computer (Figure 3). During measurements, participants refrained from talking or using electronic devices to minimize interference. Firstly, EEG data were preprocessed with a 0.5–100 Hz band-pass and 50 Hz notch filtering, re-referenced to the common average, which is as follows:

$$y_t={\displaystyle \sum }_{k=0}^M{b}_k{x}_{t-k}-{\displaystyle \sum }_{k=1}^N{a}_k{y}_{t-k}$$

where x_t is raw EEG data, y_t is filtered EEG data, and b_{k and} a_k are filter coefficients. Subsequently, EEG waveforms contaminated by signal interference or artifacts resulting from frequent eye blinks, head movements, or other participant actions were removed. EEG artifacts were isolated and removed using independent component analysis (ICA). Artifactual components were identified according to standardized criteria, including excess kurtosis (k > 5) indicating non-Gaussian artifact signatures, high spatial correlation (r > 0.7) with canonical EOG and EMG topographies reflecting ocular/muscular contamination, and dominant spectral power concentration in artifact-specific frequency bands. Then, EEG frequencies are categorized into five bands: delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (14–30 Hz), and gamma (>30 Hz). Alpha waves (8–13 Hz) indicate relaxation states, while beta waves (14–30 Hz) reflect cognitive engagement [51]. Frequency band ratios effectively reveal relative neural activity changes across bands, offering nuanced insights into stress recovery dynamics [52,53]. Specifically, the α/β ratio serves as a validated indicator where higher values correspond to relaxed states [54]. Specifically, EEG frequencies are extracted and classified by the following formula:

$$F\left(\omega \right)={\int }_{-\infty }^{\infty }f\left(x\right)e^{-i\omega x}dx.$$

In these formulas, f(x) denotes the raw time domain EEG signal, where t is time. The e^−iωx is the rotation factor employed to extract the signal component at the specific angular frequency ω. Finally, we calculated power in the frequency scale from 0.1 Hz to 30 Hz. Grand averages of aperiodic exponents and band oscillations were calculated from pooled electrode locations in various regions of interest (ROIs): left frontal (F3, F7, AF3, and FC5), right frontal (F4, F8, AF4, and FC6), occipital (O1, 02), and parietal (P7, P8). To calculate the average power within a specified frequency band [f₁,f₂] and calculate the α/β ratio, use the following formulas:

$$P_{mean}\left(c\right)={\displaystyle \frac{1}{B}}{\sum }_{i\in band}P\left(i,c\right)$$

$${\displaystyle \frac{\alpha }{\beta }}={\displaystyle \frac{P_a}{P_b}}.$$

In these formulas, P is the power data array of array of dimensions frequency points*channels, band contains indices of the frequency points within the band [f₁,f₂], B is the number of frequency points in the band, P(i, c) is the power value at frequency index i and channel c, and P_mean(c) is the mean power for channel c across the specified frequency band. P_a is mean power spectral density in the α band, and P_b is mean power spectral density in the β band. EEG data (Figure 4) were processed using the EEGLAB plugin within MATLAB 2022 software.

3.3.3. Measurement of Psychological Indicators

The questionnaire consisted of several parts to measure perceived environmental perception, perceived restorativeness, and psychological restoration. The selection of these dimensions considered the independence between dimensions and their widespread acceptance in both academic and practical domains.

For landscape perception, three questions were adapted from previous studies to measure the perceived qualities of the landscape, i.e., naturalness, beauty, and complexity [35,55,56]. Item consistency was tested (Cronbach’s α = 0.821). Our evaluation of spatial perception focused on two aspects, using the indicators of “openness” and “safety” [29]. The results show high internal consistency for these two dimensions (Cronbach’s α = 0.798). For facility perception, the indicators of “completeness” and “accessibility” were used. Item consistency was tested (Cronbach’s α = 0.902). The biodiversity perception evaluates participants’ perceptions across four dimensions: plant color diversity, plant species diversity, bird sound diversity, and bird species diversity. The results show high internal consistency for these four dimensions (Cronbach’s α = 0.837). All perception scales were designed by the research team on a scale from 1 to 7, where 7 represented the highest positive rating (agree strongly) and 1 represented the lowest rating (disagree strongly).

Drawing on research in psychology and landscape ecology, we observed that previous studies developed tools for assessing long-term psychological states, such as the General Health Questionnaire and the Mental Health Scale [57]. However, these tools are unsuitable for our study, which aims to explore the immediate psychological benefits following park visits. Consequently, we consulted several relevant studies and, using their methodology, developed five items designed to measure the impact of various landscapes on young adults’ immediate psychological health benefits across different environments (

Table 1. Summary of outcome construct and perceptual factors.

Construct	Item		Reference
Mental restoration	MR1: How would you rate the improvement in self-perceived energy levels after this visit? MR2: How would you rate the improvement in self-perceived health status after this visit? MR3: How would you rate the improvement in self-perceived confidence after this visit? MR4: To what extent do you feel that this visit relaxed you? MR5: To what extent do you feel that this visit restored your mood?		[58] H. Liu et al. (2017); [59] Zhou et al. (2022)
Perceived restorativeness	Fascination	F1: Places I like that are fascinating. F2: My attention is drawn to many interesting things. F3: I feel really drawn to details in this place. F4: I would like to spend more time looking at the surroundings.	[61] Hartig et al. (1997)
	Being away	BA1: There is a different vibe here. BA2: I feel really detached from my daily routine. BA3: Being here is an escape experience. BA4: I can relax here. BA5: I feel really detached from the stress of everyday life.
	Extent	E1: I am confused here. E2: There is too much going on. E3: There is a great deal of distraction here.
	Compatibility	C1: I have a sense that I belong here. C2: Being here suits my personality. C3: I can focus on my activities without interruptions.

) [58,59,60].

Additionally, the Perceived Restorative Scale (PRS) was employed as a mediating variable to investigate the influence of environmental perception on physiological and psychological outcomes. The PRS was used based on a seven-point Likert scale, which comprises four dimensions that assess the restorative qualities of landscapes, offering an in-depth understanding of how environmental perception influences individual well-being. As there is evidence that condensed dimensions could be available [61], this study adopted an abbreviated 15-item PRS, in which fascination was measured with 5 items. Being away was measured with 5 items. Extent and compatibility were measured with 3 items. The results show high internal consistencies for restorative dimensions (Cronbach’s α = 0.787).

3.3.4. Experimental Procedure

Upon participants’ arrival at the study site, the experimenter provided detailed explanations of the research goals, procedures, and the use of EEG equipment to ensure that participants understood clearly. Participants were then asked to observe the landscape at four designated observation points, either sitting or standing in sequence, while EEG signals were recorded continuously during each session. Before starting EEG data collection, participants completed a brief stress induction test to standardize their initial emotional and stress levels. This test involved solving mathematical problems of increasing difficulty. Afterward, participants engaged in a landscape perception phase lasting about 10 min at each location, during which EEG data were continuously recorded until the perception period ended. After completing the perception phase at each site, participants were asked to fill out three questionnaires: the Park Perception Scale, PRS, and the Mental Restoration Assessment Questionnaire (Figure 5).

3.4. Statistical Analyses

To validate the objective drivers of environmental perceptions, perception responses were modeled using linear mixed-effects models (LMMs) to address nested data structure (60 participants*4 sites). Four environmental types were treated as a random intercept to control for unmeasured site-level heterogeneity. Fixed effects included quantified landscape elements. The perception scores were modeled as follows:

$$Y_{ij}{=\beta }_0+{\beta }_1{X}_{1j}+···+{\beta }_p{X}_{pj}+{\mu }_j+{\epsilon }_{ij}.$$

where Y_ij is the perception score of the ith individual in the jth site, X_Pj are site-level landscape metrics, μ_j is the random intercept for different sites, and ε_ij is the individual-level residual. Significant predictors (α = 0.05) identified by likelihood ratio tests were then visualized via linear regression (GLMs) using scene-averaged perception scores. All predictors were standardized to a zero mean and unit variance (Z-scores), as is recommended practice when working with predictors on different scales. All analyses used REML estimation in R 4.4.1 (lme4 and lmerTest packages).

To quantify restorative benefits across environmental scenarios, we conducted a one-way analysis of variance (ANOVA) on physiological data (EEG-α/β ratio) and psychological data (PRS scores). Tukey’s HSD post-hoc multiple comparisons were subsequently applied to identify which specific landscape settings exhibited significant advantages in terms of psychological or physiological restoration. Moreover, differences in environmental perception variables across the four experimental sites were analyzed using the Kruskal–Wallis H test, a non-parametric alternative to ANOVA appropriate for ordinal data or when normality assumptions are violated. This test was selected due to the Likert scale nature of perception variables and heterogeneous variances between groups. Post-hoc pairwise comparisons with Dunn–Bonferroni adjustment were performed for variables showing significant omnibus effects (p < 0.05). The ANOVA and Kruskal–Wallis H tests were conducted in Excel 2016 and SPSS Statistics 23.0.

To explore the mediating mechanisms of perceptual pathways, structural equation modeling (SEM) was employed. SEM was chosen for its capacity to model multiple dependent variables and specified mediating effects within a unified framework, aligning with our hypothesized pathways [62]. The SEM was employed in AMOS 25.0.

4. Results

4.1. Relationships Between Landscape Elements and Perceptual Dimensions

In the current study, linear mixed-effects models identified significant environmental predictors controlling for site-level clustering (ICC = 0.24), informing subsequent linear regression visualizations. In the current study, linear regression analysis demonstrated a statistically significant correlation between environmental elements and perceptual dimensions (Figure 6). Landscape perception showed positive associations with landscape richness, waterbody proportion, and plant diversity, suggesting that visually complex natural features enhance scenic engagement through multisensory stimulation. Spatial perception is negatively associated with vegetation proportion, but positively associated with waterbody and sky proportions, indicating that expansive features strengthen prospect-dominated experiences. Biodiversity perception was positively affected by bird diversity, landscape richness, and plant diversity, indicating that auditory and visual cues from fauna and flora jointly shape the perception of wildlife. However, facility perception exhibited insignificant correlation with environmental elements, indicating that perceptions of facilities are primarily driven by their intrinsic attributes (quality, quantity, and maintenance) rather than visual integration with surrounding natural features. Collectively, these model-validated linkages confirm that perceptual assessments are grounded in quantifiable environmental properties, establishing a foundation for subsequent mediation analyses of restoration pathways.

4.2. Comparison of Restorative Benefits Across Different Study Sites

The results of repeated-measures ANOVA revealed distinct patterns in both physiological and psychological responses across the different environmental settings (Figure 7, Table S1), indicating that various types of wetland park environments can induce changes in brain activity and psychological restoration. The physiological data, quantified by the EEG-α/β ratio, showed a ranking of median values from high to low as waterfront, forest, wetland, and plaza. However, significant differences were only observed between the waterfront and plaza (p < 0.05). These results suggest that in the waterfront environment, the parasympathetic nervous system is most active, corresponding to the lowest cognitive load and significant alleviation of stress and anxiety. In contrast, the plaza environment exhibited the lowest median EEG-α/β ratio, indicating that participants experienced a significant cognitive load and high levels of stress.

The psychological data, derived from mental restoration scores, revealed distinct variations among the four study environments. The plaza environment exhibited the lowest median score, with significant differences observed between plaza, waterfront/wetland environments (p < 0.001), and between plaza and forest (p < 0.01). These findings indicate that the plaza environment is associated with less favorable mental health outcomes compared to waterfront, forest, and wetland settings.

The Kruskal–Wallis H test results, displayed in

Table 2. Differences in perception variables in four study sites.

Perception Variables	Types of Experimental Sites (Mean ± SE)				Test Statistics
	Waterfront	Plaza	Forest	Wetland	F	p
Landscape perception
Naturalness	4.93 ± 0.31	4.56 ± 0.30	5.37 ± 0.27	5.56 ± 0.23	2.525	0.058
Beauty	1.84 ± 0.30	1.43 ± 0.27	2.44 ± 0.27	2.53 ± 0.27	3.502	0.016 *
Complexity	3.71 ± 0.31	3.47 ± 0.31	4.57 ± 0.30	3.97 ± 0.29	2.419	0.067
Spatial perception
Openness	3.35 ± 0.28	2.66 ± 0.29	2.41 ± 0.29	2.74 ± 0.31	1.948	0.093
Safety	2.64 ± 0.32	2.47 ± 0.26	2.28 ± 0.30	3.29 ± 0.30	2.862	0.038 *
Facility perception
Completeness	3.56 ± 0.27	2.83 ± 0.29	3.29 ± 0.30	3.46 ± 0.30	1.199	0.311
Accessibility	3.75 ± 0.29	3.25 ± 0.30	3.63 ± 0.30	3.98 ± 0.31	0.997	0.395
Biodiversity perception
Plant colors	3.69 ± 0.29	2.90 ± 0.29	4.24 ± 0.27	4.29 ± 0.31	4.879	0.003 **
Plant species	5.00 ± 0.25	3.86 ± 0.29	5.27 ± 0.25	5.10 ± 0.28	5.683	0.001 *
Bird sounds	4.17 ± 0.30	3.08 ± 0.27	4.49 ± 0.30	4.44 ± 0.35	3.807	0.011 *
Bird species	4.42 ± 0.30	2.93 ± 0.34	4.41 ± 0.32	4.51 ± 0.37	6.164	0.000 ***

* Denotes statistically significant difference (* p < 0.05; ** p < 0.01; and *** p < 0.001).

, show significant variation in young adults’ landscape, spatial, and biodiversity perceptions across the four sites. Regarding landscape perception, the “beauty” attribute achieved its highest scores in the wetland environment and its lowest in the plaza. For spatial perception, the wetland environment was strongly linked to the perception of safety, whereas the forest was perceived as the least safe area. Concerning biodiversity perception, the plaza exhibited the lowest scores for plant colors, plant species, bird sounds, and bird species. In contrast, the wetland environment had the highest values for plant colors and bird species, and the forest area recorded the highest values for plant species and bird sounds. Yet, other perception variables showed no significant differences among the groups.

4.3. The Linkage Between Wetland Park Perception and Psychological and Physiological Restoration

Two structural equation models were built based on the experimental design. Model 1 examined the association between wetland park perceptions and physiological data (EEG-α/β ratio). Model 2 examined the correlation between wetland park perceptions and psychological data (mental restoration). The goodness-of-fit indices of the proposed concept model were estimated by the maximum likelihood method [63], including four absolute goodness-of-fit indices, i.e., χ²/df, comparative fit index (CFI), goodness-of-fit index GFI, root mean square error of approximation (RMSEA), and one relative goodness-of-fit index, the incremental fit index (IFI). The results are shown in Table 2. All the indices are with values in the range of the respective recommended values, which means a good goodness-of-fit of the models.

The results of Model 1 (

Table 3. The fitting index of the model.

	Dependent Variables	χ2/df (<3)	RMSEA (<0.08)	GFI (>0.9)	IFI (>0.9)	CFI (>0.9)
Model 1	EEG-α/β ratio	1.465	0.044	0.933	0.968	0.977
Model 2	Mental restoration	1.607	0.051	0.921	0.977	0.967

, Figure 8, and Table S2) show that landscape perception, spatial perception, and facility perception had significant positive effects on the EEG-α/β ratio (p  <  0.001). Conversely, the path from biodiversity perception to EEG data did not show a significant effect (p = 0.334). For the aspect of PRS, landscape perception, spatial perception, facility perception, and biodiversity perception had significant positive effects on PRS. Moreover, PRS was positively associated with the EEG-α/β ratio (p  <  0.001).

The results of Model 2 (Table 3, Figure 8, and Table S3) show that landscape perception, spatial perception, and biodiversity perception had significant positive effects on mental restoration (p  <  0.001), and the path from facility perception to mental restoration (p = 0.326) was not significant. For the aspect of PRS, landscape perception, spatial perception, facility perception, and biodiversity perception had significant positive effects on perceived restorativeness. Moreover, PRS was positively associated with mental restoration (p  <  0.001).

4.4. The Mediating Effects of PRS on EEG-α/β Ratio and Mental Restoration

To further explore the mediating effects of perceived restorative settings (PRS), we analyzed the indirect impact of PRS on physiological and psychological outcomes through both the EEG-α/β ratio and mental restoration (

Table 4. Total, direct, and indirect effects of various restorative potentials on the dependent variables of EEG-α/β ratio and mental restoration (asterisks ** and *** indicate significance levels of p < 0.01, and p < 0.001).

Dependent Variables	Linkages	β Unstandardized (95% CI)	p-Value
EEG-α/β ratio	Landscape perception→EEG-α/β ratio
	Total effect	0.28 **(0.243, 0.037)	<0.01
	Direct effect	0.243 ***(0.2, 0.286)	<0.001
	Indirect effect	0.037 **(0.019, 0.055)	<0.01
	Spatial perception→EEG-α/β ratio
	Total effect	0.371 ***(0.314, 0.057)	<0.001
	Direct effect	0.314 ***(0.269, 0.395)	<0.001
	Indirect effect	0.057 **(0.034, 0.081)	<0.01
	Facility perception→EEG-α/β ratio
	Total effect	0.298 ***(0.223, 0.075)	<0.001
	Direct effect	0.223 ***(0.186, 0.26)	<0.001
	Indirect effect	0.075 ***(0.057, 0.093)	<0.001
	Biodiversity perception→EEG-α/β ratio
	Total effect	0.081 ***(0.059, 0.022)	<0.001
	Direct effect	/	/
	Indirect effect	0.022 ***(0.008, 0.036)	<0.001
Mental restoration	Landscape perception→Mental restoration
	Total effect	0.208 ***(0.3, 0.06)	<0.001
	Direct effect	0.155 **(0.264, 0.41)	<0.01
	Indirect effect	0.053 **(0.111, 0.017)	<0.01
	Spatial perception→Mental restoration
	Total effect	0.248 **(0.360, 0.135)	<0.01
	Direct effect	0.190 **(0.310, 0.097)	<0.01
	Indirect effect	0.058 **(0.117, 0.019)	<0.01
	Facility perception→Mental restoration
	Total effect	0.159 ***(0.259, 0.059)	<0.001
	Direct effect	/	/
	Indirect effect	0.159 ***(0.259, 0.059)	<0.001
	Biodiversity perception→mental restoration
	Total effect	0.323 **(0.412, 0.197)	<0.01
	Direct effect	0.220 ***(0.308, 0.078)	<0.01
	Indirect effect	0.103 ***(0.176, 0.048)	<0.001

). The mediation analysis revealed distinct patterns of partial and complete mediation in the relationship between environmental restorative potentials and these outcomes. In terms of physiological outcomes, PRS perception partially mediated the relationships between landscape perception, spatial perception, facility perception, and the EEG-α/β ratio. Path analysis revealed a strong direct effect of landscape perception and spatial perception on the EEG-α/β ratio (β = 0.243 and 0.314, p < 0.001). Although the indirect effect via perceived restorativeness (PRS) was statistically significant (β = 0.037 and 0.057, p < 0.01), its magnitude was substantially weaker than the direct pathway. In contrast, biodiversity perception operated exclusively via complete mediation for the EEG-α/β ratio (β = 0.081, p < 0.001), relying entirely on RPS-linked appraisals. Regarding psychological outcomes, PRS perception partially mediated the pathways from landscape perception, spatial perception, and biodiversity perception to mental restoration. However, the impact of facility perception on mental restoration was found to be fully mediated by PRS perception (β = 0.159, p < 0.001).

5. Discussion

5.1. Different Environments Modulate Restoration Efficacy and Perceptual Preferences

This study investigated physiological and psychological restoration among young adults across four wetland park environments, with an emphasis on elucidating how perceptual preferences drive differential restorative outcomes. There is evidence indicating that various green space exposures have distinct effects on the recovery of positive emotions [11,64]. Similarly, we demonstrated that 10 min exposures to distinct environments elicited significantly varied recovery patterns. Regarding physiological restoration, young adults exposed to waterfront areas exhibited a significantly improved EEG-α/β ratio compared to those in plaza environments. This finding aligns with Ulrich’s stress reduction theory, which posits that aquatic elements have innate stress-reducing properties by activating parasympathetic pathways [46]. Jin et al. (2024) also recently reported that blue spaces have the most efficient thresholds of EEG compared to open green spaces, semi-closed green spaces, and grey spaces [53]. Notably, physiological and psychological restoration exhibited distinct environmental patterns, and psychological restoration demonstrated broader efficacy, with wetlands, waterfronts, and forests collectively outperforming plazas at statistically significant levels. This might be because physiological recovery is primarily driven by direct stimuli, whereas psychological recovery is based on cognitive evaluation, leading to different sensitivities of the two environmental characteristics [48,65]. The differentiation stems from distinct cognitive processing modes: the direct sensory-driven recovery operates via subcortical pathways, where visual–auditory stimuli (e.g., water movement, light reflections) trigger autonomic nervous system responses within seconds, bypassing conscious appraisal. This aligns with Ulrich’s SRT, emphasizing nature’s pre-attentive processing. Cognitively mediated recovery engages prefrontal appraisal systems, requiring top-down evaluation of environmental meaning over several minutes [46,66]. These results are also consistent with previous studies indicating that a natural setting has a significant positive effect on mental restoration compared to impervious areas [67,68]. Moreover, the plaza’s lowest EEG-α/β ratio and psychological restoration scores might be attributed to the absence of natural features such as water bodies and abundant vegetation, corroborating the literature on the cognitive burdens and “attentional fatigue” of impervious space [25,49].

This study revealed that young adults’ perceptual preferences varied significantly across wetland park environments, potentially modulating their physiological and psychological restoration outcomes in these settings. In terms of landscape perception, the wetland area scored significantly higher than the plaza area in terms of the “beauty” attribute. This distinction can be attributed to the dynamic interplay of water features and vegetation in wetlands, which creates rich visual layers [69]. Such a landscape configuration aligns with the prospect-refuge theory, where open water provides prospects and surrounding vegetation offers shelter. In contrast, plaza areas, characterized by limited natural diversity and monotonous structures, fail to evoke positive emotional responses. Regarding spatial safety perception, wetlands were perceived as the safest environment, while forests were considered the least safe. This difference highlights the key role of visual permeability. Wetlands’ semi-open features, combining open water views with vegetation boundaries, allow young adults to quickly assess their surroundings and meet defensible space needs. Conversely, forests’ dense vegetation obstructs sightlines, causing concerns about potential threats [70]. In terms of biodiversity perception, the plaza area exhibited the lowest values across multiple dimensions, including plant colors, plant species, bird sounds, and bird species. This can be explained by plazas’ lower species diversity, simpler ecological functions, and lack of natural sensory engagement and immersive experiences [71]. Young adults often accumulate prior knowledge about natural ecosystems during their long-term learning process. This foundational knowledge enables them to more readily recognize and comprehend the ecological roles of organisms within these environments when directly experiencing them, thereby fostering more pronounced biodiversity perceptions [72].

5.2. Multi-Perception Drive Dual-Pathway Restoration

The mediation analysis of SEM reveals distinct pathways through which wetland park environments foster restoration in young adults, demonstrating how physiological and psychological outcomes arise from different combinations of direct perception and cognitive appraisal. For physiological restoration measured by EEG-α/β ratio, sensory-driven perceptions, including landscape, spatial, and facility perception, exert strong direct influences on stress reduction, which is consistent with stress reduction theory [16]. These elements appear to trigger innate neurophysiological calming responses independently of conscious evaluation. These observations align with previous research suggesting that exposure to an environment with different levels of openness, landscape richness, and naturalness leads to distinct fluctuations in brain activities [52,53,73]. Moreover, the current study first explores the association between biodiversity perception and brain activity using SEM. Our findings indicate that biodiversity contributes to physiological restoration only through the mediating role of PRS, indicating that its calming effect requires conscious recognition of the environment’s restorative qualities rather than operating through automatic sensory pathways. The restorative benefits of biodiversity perception may exhibit a delayed effect: complex ecological information might initially increase cognitive load (e.g., during species identification attempts), temporarily diminishing relaxation effects [11,74]; once PRS appraisal is established, its restorative benefits manifest in EEG through neuroendocrine regulatory pathways.

Psychological restoration follows a similar pattern. Landscape, spatial, and biodiversity perceptions directly enhance mental restoration while concurrently operating through the mediation of perceived restorative settings (PRS), indicating dual restorative mechanisms: an immediate experiential pathway and a cognitive–reflective pathway. Notably, biodiversity perception demonstrates significant direct psychological value despite its lack of direct physiological impact, highlighting its distinct role in conscious restoration processes [75]. Our findings are in line with previous work that tested the direct link between biodiversity and mental well-being [76]. The perception of biodiversity fosters an appreciation for the value of life, leading to an immediate enhancement of self-reported restorative experiences [77]. In contrast, facility perception influences psychological outcomes exclusively via PRS mediation. This implies that infrastructural elements (e.g., pathways, benches) support restoration not through inherent properties, but by enabling environmental appraisals conducive to reflection and recovery.

Overall, young adults’ perceptual preferences generate physiological and psychological restoration through two complementary theoretical frameworks: a direct sensory-driven stress reduction mechanism and a cognitively mediated attention restoration pathway. These findings indicate that wetland parks deliver layered restorative experiences where sensory elements directly reduce physiological arousal, whereas ecological complexity and functional design necessitate cognitive engagement to fully realize their restorative potential.

5.3. Implications for the Design and Planning of Urban Smart Parks

The empirical findings from this study offer valuable strategies for embedding smart restorative technologies into urban parks, positioning them as dynamic health infrastructures within sustainable built environments (Figure 9). The direct sensory drive pathway is mainly characterized by fast, low-order sensory neural coupling, manifested by a rapid increase in EEG-α/β values within 10 min; the triggering threshold depends on instantly visible/audible “low complexity natural cues” [78]. Waterfront environments consistently demonstrated superior restoration potential, as evidenced by significantly enhanced EEG-α/β ratios after brief exposures. To maximize restorative efficacy, waterfront areas should be prioritized as key restorative zones to create innately processed stimuli. This can be achieved through real-time biofeedback systems and adaptive microclimate controls (lighting/soundscapes activated by motion sensors), complemented by expanded aquatic plantings monitored via soil moisture sensors. Complementarily, forest habitats require perceptual augmentation to address safety concerns stemming from visual obstruction. LiDAR-informed vegetation management to enhance vertical visibility, coupled with elevated boardwalks providing overall perspectives, can mitigate perceived threats while preserving ecological integrity [79]. For underperforming impervious plazas, interventional greening should integrate biodiverse pocket wetlands with audible water features to counteract cognitive burdens inherent to hardscape.

Critically, biodiversity’s exclusive reliance on the cognitive mediation pathway requires individuals to re-evaluate the meaning of the environment, using the PRS score as a mediating variable and demanding 3–5 min of cognitive integration time. Design strategies, therefore, shift toward technology-aided ecological legibility. However, the exclusive mediation of biodiversity effects via PRS also signals a potential mismatch between actual ecological richness and visitors’ conscious detection. This awareness gap suggests that parks with high species value may still underperform in restoration unless ecological information is translated into legible experiences. To close this gap, we recommend embedding low-cost, non-intrusive interpretive layers [80]. For instance, infrared-triggered camera walls strategically positioned along trails enable real-time observation of animal activities, while glass viewing panels embedded in boardwalks expose underwater ecological processes. Furthermore, designing seasonal variations in plant colors can enhance public awareness of natural cycles and enrich their perception and experience of nature. Tiered interpretation systems that connect ground-level plant markers to canopy-level avian acoustic displays through QR codes create layered learning opportunities for understanding ecological processes. These approaches address biodiversity’s delayed restorative benefits by reducing initial cognitive load during species identification.

Third, infrastructures could function as restoration catalysts rather than standalone elements. Landscape facilities should integrate multi-sensory elements to strengthen perceptual engagement. Visually diverse plantings and landform features can create aesthetically rich experiences that stimulate both sight and sound. Spatial layouts integrate open vistas with intimate enclosed spaces through interconnected curvilinear pathways and vertical layering, thereby enhancing environmental immersion. To enhance biodiversity understanding, facilities can incorporate ecological habitats alongside interpretive features such as wildlife corridors. Modular platforms supporting temporary installations such as field microscopes exemplify this approach, transforming infrastructure into cognitive bridges between users and natural systems.

Ultimately, a dynamic closed-loop framework that integrates environmental sensors, user biofeedback, and adaptive systems was proposed. To bridge the implementation gap between theoretical proposals and municipal practice, our smart park framework advocates phased adoption anchored in existing urban infrastructures: (1) cost-efficient sensor integration by retrofitting environmental monitors onto streetlights/park facilities rather than new installations; (2) policy–priority alignment targeting districts with pre-allocated “sponge city” or mental health initiative budgets to leverage synergies; and (3) scalable pilots initiating biofeedback zones in high-traffic areas before park-wide rollout, using revenue from tourism uplift to fund expansion.

5.4. Limitations of the Study

Although this study provides novel insights into wetland restoration pathways from the physiological and psychological perspectives of young adults, some limitations of this study should be noted as well. (1) This experiment chose autumn as the experimental period and did not consider the effect of season on the results. Future studies should include seasonal variables, conduct comparative experiments, and explore other potential factors that may affect the restorative outcomes. (2) Physiological measurements relied solely on EEG-α/β ratios as indicators of cortical arousal, neglecting complementary biomarkers such as heart rate variability (HRV) or salivary cortisol that could offer multidimensional validation of stress recovery. (3) The 10 min exposure protocol, though standardized for cross-environment comparison, may not capture longer-term restorative dynamics; circadian effects and cumulative benefits of repeated exposures remain unexplored. Due to the high complexity of the natural environment, it is necessary to further focus on the analysis of longitudinal time series. Future research on multiple spatiotemporal scales is a necessary condition for analyzing the mechanisms of environmental characteristics on mental health.

Meanwhile, this study also recognizes several practical issues that smart park development may face: (1) the collection and processing of EEG-oriented biofeedback, video/acoustic recordings, and pedestrian heat map data may expose personally identifiable and sensitive information, creating privacy and security risks. Future research can focus on the planning level by defining data collection boundaries and purposes, designating no-collection zones for sensitive areas, instituting prominent notice and public consultation, and implementing data minimization and periodic evaluations to improve smart park development; and (2) elevated boardwalks, interactive displays, and QR-code-based interpretation may be inaccessible or impractical for vulnerable groups and users with low bandwidth or without devices, raising concerns about fairness and accessibility [81]. Future research and design should further implement accessibility and universal design—for example, continuous ramps and alternative paths, rest nodes, physical wayfinding/multilingual signage, and offline services—to reduce dependence on digital devices.

6. Conclusions

Situated within a social–ecological framework, this empirical study confirms the measurable mental health benefits of wetland parks for young adults. We found that different habitats elicit varied restorative outcomes: waterfront areas triggered the strongest physiological recovery (significantly elevated EEG-α/β after brief exposure), whereas impervious plazas consistently underperformed, underscoring the essential role of natural features. Crucially, restoration operates through dual pathways—sensory-driven landscape and spatial perceptions directly reduce physiological stress, while biodiversity contributes to physiological restoration solely via PRS mediation. For psychological restoration, landscape, spatial, and biodiversity perceptions act both directly and via PRS mediation, whereas facility perception operates exclusively through PRS. Building on these insights, we propose actionable smart park design strategies. This framework offers transferable solutions for river network megacities globally, providing blueprints to transform green spaces into intelligent infrastructures that mitigate urban stress. Future research should (1) conduct cross-cultural comparative studies across diverse wetland parks (e.g., coastal vs. inland, temperate vs. tropical) to disentangle ecological and cultural influences on restoration pathways; (2) prioritize co-design with communities to address perceptions of technology intrusiveness, privacy concerns with biofeedback, and equitable access.