Quantifying Spatiotemporal Characteristics of Urban Wetland Soundscapes and Their Associative Pathways Regulating Restorative Benefits

Abstract

The soundscape serves as a critical determinant of the quality of urban wetland parks. This study employs a mixed-methods approach to comprehensively evaluate wetland soundscapes. First, field investigations combining sound level measurements and questionnaire surveys were conducted in Aixi Lake Wetland Park to analyze the spatiotemporal characteristics of the soundscape. Second, laboratory-based physiological tracking (using wearable sensors) and cognitive tests (Sustained Attention to Response Task, SART) were utilized to experimentally quantify the restorative benefits of typical soundscapes. The findings reveal that: (1) sound level indicators and sound harmonious degree in urban wetland parks exhibit significant spatiotemporal characteristics and distributional variations; (2) a marked competitive effect among biological, geophysical, and human activity sounds is observed in their spatial distribution; sound harmonious degree demonstrates significant spatial autocorrelation in both global and local models; (3) different sound sources possess varying restorative potentials, with bird song showing the highest restorative effect; the SHDs of biological and geophony, along with L_Aeq, are key factors affecting PRSS; (4) a positive correlation exists between L_Aeq and the PRSS up to 56.4 dB, beyond which PRSS declines with increasing L_Aeq; (5) at the physiological level, short-term exposure to urban wetland park soundscapes can rapidly alleviate stress, with the most pronounced restorative effects occurring within the first 60 s; and (6) in terms of attention, soundscape stimulation reduces SART response times and improves response speed, while bird song from treetops and musical sounds further decrease response errors.

1. Introduction

Rapid urbanization has significantly amplified the psychological stress of daily life, increasingly contributing to chronic physical and mental health disorders [1]. Urban green spaces (UGS) help mitigate these adverse effects by providing restorative microclimates that actively alleviate anxiety and mental fatigue [2]. Broadly defined by the United States Environmental Protection Agency (EPA) as vegetated land areas—encompassing public parks, community gardens, and greenways—these environments deliver essential ecosystem services, ranging from air purification and urban heat island reduction to psychological restoration. Among these, urban wetland parks function as a particularly vital component of this green infrastructure.

Natural environments have long been recognized for their capacity to promote physical and mental well-being. Building on this foundation, Kaplan formalized the concept of the “restorative environment” to describe spaces that actively alleviate psychological fatigue and stress [3]. Historically, however, environmental perception research has been heavily biased toward visual landscapes [4,5,6]. The auditory environment—the second major sensory channel for processing spatial information—offers equally vital, though frequently overlooked, restorative benefits [7,8]. The concept of the “soundscape,” pioneered by R. Murray Schafer in the 1970s and later standardized by ISO 12913-1:2014, shifts the focus from purely objective noise metrics to the acoustic environment as perceived and understood by people in context [9]. This definition establishes a core tripartite framework comprising the human listener, the physical space, and the sound itself [10]. Consequently, modern soundscape studies demand an interdisciplinary approach that integrates landscape architecture, psychoacoustics, and environmental acoustics to fully capture these complex relationships (Figure 1). While empirical soundscape ecology is a relatively modern scientific discipline, the esthetic and cultural appreciation of acoustic environments has deep historical roots. In China, intuitive explorations of soundscape perception date back to the Book of Songs (11th–6th century BC), where early poetic documentations frequently highlight the emotional and spatial resonance of environmental sounds [11]. This enduring cultural legacy underscores the profound, inherent connection between humans and their acoustic surroundings.

Mapping the spatiotemporal attributes of acoustic environments is a central focus of modern soundscape management [12]. While early spatial studies, such as those by Bai et al. [13], successfully mapped urban noise pollution, they primarily treated sound as a physical hazard rather than exploring subjective human perception. Subsequent research by Wu and Chen [14,15] advanced the field by tracking spatiotemporal soundscape dynamics within historical districts and urban parks; however, these studies stopped short of linking acoustic distribution to physiological or psychological stress recovery. Although the theoretical framework for soundscape restoration is still evolving, recent efforts have established robust assessment criteria for healthy environments. For example, Kogan et al. [16] developed the Health Restoration Soundscapes (HeReS) conceptual model to evaluate urban green spaces. Empirical evidence increasingly demonstrates the health benefits of natural sounds, with researchers like Lam et al. [17] successfully utilizing natural audio in masking systems to mitigate urban traffic noise. Furthermore, subjective evaluation tools—such as the Perceived Restorativeness Soundscape Scale (PRSS)—have confirmed that urban green spaces deliver quantifiable restorative benefits [18,19]. These benefits, however, are heavily modulated by individual differences [20,21]. For instance, frequent park visitors report higher perceived restorativeness [18], and demographic factors shift acoustic preferences; while adults generally find natural sounds most restorative, children often prefer music or singing [22]. Additionally, Jeon et al. [23] demonstrated that acquired experiential traits influence soundscape perception more significantly than innate personality profiles. Despite growing recognition of the acoustic restorative capacity of urban parks [24,25], a critical literature gap remains: no systematic research has yet correlated the spatiotemporal dynamics of these soundscapes directly with their physiological and cognitive restorative benefits. Addressing this gap is essential for evidence-based urban planning. By mapping spatiotemporal acoustic data and experimentally quantifying the restorative benefits of different soundscapes, this study aims to provide actionable design strategies for optimizing the acoustic environments of urban wetland parks, ultimately promoting the physical and mental well-being of urban residents.

Addressing this research gap directly supports global sustainable development, specifically the United Nations Sustainable Development Goal 11 (SDG 11) for inclusive, safe, and resilient cities [26]. As core components of urban green infrastructure, wetland parks are essential for delivering simultaneous co-benefits to both ecological conservation and public health [27]. However, current ecosystem service assessments disproportionately prioritize provisioning and regulating functions, such as hydrological control and biodiversity maintenance. This traditional focus systematically neglects soundscape-mediated cultural ecosystem services—most notably, the provision of intangible restorative and mental health benefits [28,29]. This oversight hinders the integration of acoustic health metrics into broader sustainability frameworks, ultimately limiting the capacity of green infrastructure to enhance residential well-being. Furthermore, while global urban planning is transitioning toward a health-oriented paradigm [30], spatial planning for urban wetlands remains overwhelmingly visually dominant. Current design practices heavily favor visual landscape esthetics and land-use efficiency. They frequently fail to establish quantitative relationships between physical soundscape attributes, perceptual quality, and human restoration. This disconnect deprives planners of the refined acoustic zoning strategies needed to optimize restorative environments, leaving a critical evidence gap in health-focused sustainable design.

To address this gap, this study investigates the health-promoting potential of urban wetland soundscapes using Aixi Lake Wetland Park as a primary case study. Specifically, this research pursues two primary objectives: (1) to systematically analyze and quantify the spatiotemporal characteristics of the wetland soundscape, and (2) to determine the associative relationships and underlying mechanisms driving its restorative benefits through converging field and experimental evidence.

2. Methods

2.1. Study Area

Aixihu Wetland Park is located in the core area of the High-Tech Industrial Development Zone in Nanchang City, Jiangxi Province. It is a typical urban natural wetland. Urban natural wetlands are natural wetlands (not artificially constructed) that are formed naturally within cities (towns), at their edges or in the nearby suburbs. They belong to the transitional ecosystem between water and land [31]. Its location is in Jiangxi Province, which is one of the provinces in China with the richest wetland resources. It has Poyang Lake, an internationally important wetland, and the largest wintering ground for migratory birds in Asia, thus having a significant ecological position. The urban density in this area is high, and the land use pattern exhibits a feature of a high degree of integration of industrial, residential and ecological functions. In this context, the Aixihu Wetland Park has been planned and constructed as the core area of the city’s “green infrastructure”. Its main management goals include wetland ecological conservation, protection of migratory bird habitats, urban rainwater regulation, and provision of natural recreation and ecological education services for high-density urban areas. This study selected it as a typical case precisely because it represents, in a manner typical of economically developed and densely populated urban areas, the exploration and practice of restoring and maintaining the ecological functions of natural wetlands through moderate artificial intervention. An in-depth study of this case holds significant reference value for the optimization and sustainable management of similar urban wetlands.

The park boasts rich natural resources, including over 50,000 trees, 40,000 bamboos, and a large number of shrubs and herbaceous plants [32]. The excellent ecological environment attracts over 50 species of migratory birds [33]. The research area was divided into grids using the grid method. To enhance the accuracy of the data, the grid size should not be too large. According to the “Environmental Noise Quality Standards” (GB3096-2008) [34], after multiple field investigations and comparisons, the wetland park was finally divided into uniform square grids of 100 m × 100 m, ensuring complete coverage of the research area and setting the center of each grid as a test point (Figure 2). Based on the current situation investigation and grid division, further sample points were selected. The selection of sample points took into account the uniform distribution of functional areas, and the differences in terrain, landscape and site scale. Finally, 24 sampling points were selected in the research area, according to the principles of accessibility and typicality, and numbered for subsequent research(Figure 2).

According to the location and functional attributes of the measurement points, the study area was divided into a planned construction area (the area that is currently being planned for expansion), a humanistic landscape area (a region centered around the display of culture, art and exquisite garden landscaping), a waterfront recreational area (linear open spaces set up along the main lake shores and river courses offering a pleasant water-access experience and a chance to enjoy views), a leisure and entertainment area (the area within the park where human activities are most concentrated and commercial and service facilities are most densely located), a migratory bird conservation area (the core area with the highest ecological protection level in the park, mainly providing habitats, food sources and breeding grounds for waterfowl and migratory birds), and a forest landscape area (a region dominated by dense tree communities, forming a closed canopy space) (Figure 3).

Seventeen typical sound sources were identified within the research area. Based on the soundscape classification criteria [35] and the characteristics of the sound sources, on the basis of the three major categories of sound sources: natural sounds (i.e., geophony), biophony, and artificial sounds. The artificial sounds were further divided into human activity sounds, mechanical sounds and musical sounds (

Table 1. Typical sound source composition of the study area.

Sound Source Category	Sound Source Name
Biophony	Bird song (BS), migratory bird calls (MBC), cricket sounds (CS)
Geophony	Wind blowing through water plants (WP), rustling leaves (RL), water flow (WF)
Human Activity Sounds	Footsteps (FS), adult conversations (AC), children playing (CP), fitness activities sounds (FAS)
Mechanical Sounds	Urban traffic noise (UTN), water pump sounds (WPS), park tour vehicle noise (PTV)
Musical Sounds	Musical instrument performance (MIP), loudspeaker music (LM), phone ringtones (PR), singers’ singing (SS)

). Biophony is the collective sound produced by non-human organisms, such as birds and insects. Geophony consists of non-biological natural sounds from elements like wind, water, and thunder. The remaining categories are anthropogenic; human activity sounds stem from direct human actions like talking or walking, mechanical sounds originate from machines and technology like engines and industrial noise, and musical sounds are organized, artistic sounds intentionally created or played by people.

Simultaneously, by processing and trimming the Landsat8 remote sensing images of the study area, the Normalized Difference Vegetation Index (NDVI), a widely used remote sensing indicator, was calculated. A 10 m resolution vegetation coverage map (Figure 4) was obtained to understand the vegetation coverage situation in the study area.

2.2. Data Collection Method

2.2.1. Field Research Data

Taking into account the migration patterns of wetland birds and the characteristics of visitors, the field investigation was conducted across spring, summer, and autumn from July 2024 to July 2025. Due to cold weather and significantly reduced biological and human activities, winter was excluded from the sampling framework. It is important to clarify that the data collected across the three active seasons were pooled together for overall analysis to establish a generalized representation of the soundscape during the park’s peak operational periods, rather than being treated as separate seasonal datasets. This approach is supported by prior research indicating that while specific biological sound sources fluctuate, the overarching spatiotemporal configuration of anthropogenic noise and the resultant landscape-scale masking effects remain structurally stable in managed urban wetlands [15,36]. Statistical preliminary tests (Kruskal–Wallis) confirmed no significant deviation in spatial variance between seasons, justifying a consolidated model to focus on the generalizable restorative mechanisms of the soundscape.

Furthermore, to capture conditions of high visitor traffic, measurements were exclusively conducted during calm, sunny holidays, avoiding days with precipitation or wind speeds exceeding 5 m/s. Data collection followed the measurement standards of “Environmental Noise Quality Standards” (GB3096-2008) [34] and “Environmental Noise Monitoring Technical Specifications” (HJ640-2012) [37].

(1) Questionnaire Data

Questionnaire data collection and objective sound level monitoring were conducted simultaneously. To ensure that respondents experienced the precise acoustic environment being recorded by the instruments, and to maintain the spatial independence required for subsequent Moran’s I analyses, surveys were strictly conducted within the immediate vicinity (a 15-to-20 m radius) of each sampling point. To minimize volunteer bias and avoid convenience sampling, a systematic intercept sampling strategy was employed. For instance, approximately every third visitor crossing a predefined perimeter near the measurement station was approached. Respondents were informed of the purpose and methods of the anonymous survey in advance. A total of fifty questionnaires were distributed at each sampling point. Crucially, to control for time-of-day effects and capture continuous temporal variations, the distribution of these 50 questionnaires was evenly allocated across the four predefined measurement periods (approximately 12 to 13 questionnaires per two-hour window). This protocol ensures that the aggregated responses at a given location represent the holistic daily condition rather than being skewed by peak participation hours.

The questionnaire consisted of 3 parts. The first part investigated the basic information of the respondents, including residential type, age, education, occupation, and frequency of visits.

The second part asked respondents about their perception of sound sources identified in the preliminary survey. First, visitors are required to assess the overall noise level of the site based on their personal experience. Then, they need to evaluate three elements that determine the harmony of typical sound sources: perceived frequency (POS), perceived loudness (PLS), and preference for sound (PFS). These three elements are all derived from the acoustic scene quality evaluation system proposed by Liu Jiang et al. [15]. By calculation, the sound harmonious degree (SHD) can be obtained. The magnitude of the SHD values of each sound source indicates the quality level of the sound environment in which it is located. Finally, to identify sound sources with restorative benefits by answering the question, “Which of the sounds you hear make you feel stress-relieved and relaxed?” a five-point Likert scale was leveraged for scoring.

The third part investigated the respondents’ evaluation of soundscape restorative benefits. This study utilized the Perceived Restorativeness Soundscape Scale (PRSS) originally developed by Payne. To ensure cultural adaptability and structural validity in the Chinese context, we adopted the localized 5-dimension (Fascination, Being Away, Compatibility, Coherence, and Extensibility), 16-item PRSS structure framework previously adapted and validated by Liu et al. [38] for Chinese urban parks(

Table 2. PRSS for urban wetland parks.

Dimension	Keywords	Description
Fascination	Curiosity	This sound environment aroused my curiosity.
	Exploration	There are many things for me to explore in this sound environment.
	Interesting	My attention was drawn to many interesting sounds in this place.
Being Away	Escape	In this sound environment, I temporarily escaped from the rhythm of daily life.
	Relief	In this sound environment, I felt relieved from work pressure.
	Avoidance	This sound environment allowed me to avoid unnecessary disturbances.
Compatibility	Adaptation	I quickly adapted to this sound environment.
	Ease	It was easy to do what I wanted in this sound environment.
	Match	This sound environment matched my preferences well.
Coherence	Orderly	The sound environment here is clear and orderly.
	Belonging	The sounds I heard here were consistent with what the sound environment should be.
	Consistency	What I like to do is consistent with this sound environment.
Extensibility	Exploration	This sound environment is large enough for me to explore from different directions.
	Infinite	The scope of this sound environment seems infinite.
	Spacious	This sound environment made me feel that the space here was very spacious.

). Reliability analysis in this study showed that the Cronbach’s α coefficient for the overall soundscape restorativeness scale was 0.96, indicating excellent internal consistency that aligns with established applications of this instrument [38]. PRSS operates at the macro-level, evaluating the holistic health outcomes (alleviating mental fatigue) of the integrated soundscape based on Attention Restoration Theory (ART).

(2) Sound Pressure Level Data

All sound level measurements were conducted using integral sound level meters of Class 1 accuracy in accordance with the IEC 61672-1:2013 standard [39] (model: iSV1101, manufactured by Shengwang Company in Beijing, China). Before and after each daily on-site measurement is conducted, the sound level meter is calibrated using a 1st-level sound calibrator (model: AWAG22A, (Manufacturer: AWAG Elektrotechnik AG, Volketswil, Switzerland) which generates a calibrated sound pressure level of 94.0 dB at a frequency of 1000 Hz). The deviations of all calibration readings are all less than ±0.5 dB. If the calibration deviation exceeds this range, the data collected during that period will be regarded as invalid and the measurement will be carried out again. During the measurement, the sound level meter is fixed on a tripod, and the microphone is placed 1.5 m above the ground to simulate the height of the human ear. The microphone is pointed in the direction of the main sound source. If there is no clear dominant sound source, it is pointed upwards towards the sky, and a wind cover is installed throughout the process to minimize the interference of wind noise on the measurement results. Acoustic measurements are carried out strictly under favorable weather conditions. The specific criteria are as follows: no precipitation and wind speed lower than 5 m/s. The wind speed is monitored in real time on-site using a handheld digital anemometer (model: GM8901 (Shenzhen Jumaoyuan Technology Co., Ltd., Shenzhen, China)). If the wind speed exceeds the limit during the measurement or if precipitation begins, the measurement should be immediately stopped and resumed once the conditions are restored.

Soundscape physical parameters, including equivalent sound level L_Aeq, as well as integrated sound level indicators L₁₀ (reflecting foreground sound characteristics) and L₉₀ (reflecting background sound characteristics), were collected at 24 sampling points during the aforementioned four time periods. The synthetic parameter L₁₀ − L₉₀ was introduced to analyze the dynamic changes in sound level indicators [40].

(3) Audio Data

Using the Sony digital recording device (Sony PCM-D100), open areas in the sample sites are selected and it is installed on a tripod for fixed-point recording. The tripod height is set at 1.2 m. The microphone of the recording device should be oriented perpendicular to the wind direction to minimize interference. The single recording duration should be no less than 3 min. The audio formats collected are PCM and MP3, with dual channels at 16-bit and a sampling frequency of 44.1 kHz.

(4) Online Data: Baidu Heat Map

Baidu heat map data is based on location clustering information of smartphone users accessing Baidu products, reflecting the human traffic in a specific spatial area [41]. Fang et al.’s research on a data-driven framework for urban population dynamics further corroborated the applicability and reliability of Baidu heat map data in reflecting human traffic [42].

2.2.2. Experimental Data

(1) Pressure reduction experiment

Focusing on the physiological responses of the human body, the measurement indicators for stress reduction can be selected. The equipment used for physiological signal measurement is the EMPATICA E4 wristband physiological signal monitor (Figure 5). The collected physiological signals include skin conductance level (SCL), heart rate (HR), and skin temperature (ST). This wristband is comfortable, concealed, and non-invasive, and is widely used in laboratory and real-life measurements. The E4 wristband is connected to a smartphone or tablet via Bluetooth, allowing for real-time data viewing, easy zooming and panning to check the signals. After each test, the data will be automatically uploaded to the E4 secure platform.

(2) Attention Restoration Experiment

The Continuous Attention Test (SART) is a sophisticated and widely used classic experimental paradigm in the fields of cognitive psychology and neuroscience, specifically designed to measure and induce fluctuations in sustained attention, particularly “attentional momentary disengagement” or “absent-mindedness” states. The program was written using EPrime 2.0 software, with 112 numbers ranging from 1 to 9. Among them, there were 10 instances of the target number “5” and 100 non-target numbers. Participants were instructed to press the J key when the target number “5” appears and the F key for non-target numbers. The numbers appeared at intervals of 900 ms, and each number remained on the screen for 350 ms (Figure 6). The indicators for the cognitive test were reaction time and reaction errors. Reaction time refers to the average reaction time of the subjects to the target numbers, representing the reaction speed of the subjects; reaction errors refer to the number of times the subjects pressed the wrong J key after the presentation of the target number, representing the reaction inhibition of the subjects. The higher the scores of these two indicators, the worse the cognitive performance of the subjects. The baseline reaction time was divided into three levels: fast (≤220 ms), medium (220–270 ms), and slow (>270 ms). The baseline reaction errors were divided into two levels: fewer (≤14) and more (>15).

2.2.3. Ethical Approval and Informed Consent

All research involving human subjects in this study, including on-site questionnaire surveys and laboratory-based physiological/attention restoration experiments, was conducted in accordance with the ethical principles for social science research involving human subjects. Prior to the research implementation, all participants were clearly informed of the research purpose, research procedures, data collection content, data usage scope, and strict confidentiality measures for personal information. No personal identifiable information was collected from the participants, and all research data was only used for academic research and manuscript publication.

All participants voluntarily participated in the research and provided verbal/written informed consent (questionnaire participants provided verbal consent by filling out the anonymous questionnaire; experimental participants provided written informed consent before the experiment). The research protocol was in line with the ethical norms of the relevant academic field, and no unethical research behavior occurred during the whole process.

2.3. Data Processing Method

2.3.1. Field Data Processing

(1) Questionnaire Data Processing

After screening the questionnaire data, 1116 valid questionnaires were attained (Figure 7), with an effective rate of 93%. Reliability analysis using SPSS 29.0.2 showed the Cronbach’s α coefficient for sound source harmony evaluation was 0.83 (threshold 0.7) and for the soundscape restorativeness scale it was 0.96, indicating good questionnaire reliability.

(3) Sound Level Data Processing

The statistical methods used in this study include interpolation analysis, Spearman’s rho, and spatial autocorrelation analysis. Sound level monitoring data were analyzed via the Inverse Distance Weighting (IDW) method in ArcGIS 10.5 to create soundscape maps of various physical and perceptual indicators. IDW can convert scattered point data into continuous surface data, thereby more accurately presenting the spatial distribution of values [43]. Furthermore, studies have shown that among common interpolation methods like Kriging, spline function, and IDW, the IDW method has the highest accuracy [44].

(4) Calculation Of Sound Harmonious Degree Data

This study analyzes the spatiotemporal characteristics of soundscape indicators by introducing the SHD evaluation index proposed by Liu Jiang et al. [15] to analyze soundscape perceptual attributes. This index is determined by POS, PLS, and PFS, reflecting the consistency between the dominant position of a sound source and human preference. The formula is as follows:

$$SH{D}_{ji}=\left[1/\left(e^{\sum _{j=1}^{n}PF{S}_{ji}/n-PF{S}_{ji}}+1\right)-0.5\right]\times PO{S}_{ji}PL{S}_{ji}$$

(1)

where $j$ is the j-th sample, $i$ is the i-th sound source, and $n$ is the sample size.

(5) Inline Data Processing

Using Python 3.10, heat maps were captured every 2 h from 0:00 to 24:00 on holidays in the research area, collecting 252 images. Because of the repetitive vitality cycle of the regional characteristics, this study selected the heat map for October 1, which had the largest visitor flow according to the park’s official annual visitor flow report, to study visitor density (Figure 8).

2.3.2. Experimental Design

(1) Participant selection

To prevent mental fatigue and potential learning effects from prolonged testing, two independent cohorts of participants were recruited from local universities for the physiological measurement (Experiment 1) and the SART attention experiment (Experiment 2).

Prior to recruitment, an a priori statistical power analysis was conducted using G*Power 3.1 software to justify the sample size. For a repeated-measures within-subjects design, assuming a medium effect size (Cohen’s f = 0.25) based on previous soundscape restoration studies, an alpha level of 0.05, and a statistical power of 0.80, the minimum required sample size was calculated to be 28 [45]. Consequently, a target sample size of 30 was established for each experiment to ensure robust detection of main effects.

For Experiment 1 (Physiological Measurement), 32 eligible participants were initially recruited. After excluding 2 invalid datasets due to equipment malfunction, the final sample consisted of 30 participants (14 males, 16 females; mean age = 24.1 ± 2.8 years; comprising 15 undergraduates and 15 master’s students). For Experiment 2 (SART Test), a separate cohort of 33 participants was recruited, yielding a final valid sample of 30 (15 males, 15 females; mean age = 24.2 ± 3.2 years). All participants reported normal hearing (threshold < 25 dB), uncorrected or corrected visual acuity of ≥ 1.0, normal color vision, and no history of anxiety or cardiovascular disorders.

(2) Experimental Stimulus

The investigators used the Sony digital audio recorder (Sony PCM-D100) to collect 17 types of auditory information (Table 1). According to the questionnaire results on sound preferences, six sounds with restorative potential were selected: bird song (BS), water flow (WF), rustling leaves (RL), cricket sounds (CS), migratory bird calls (MBC), and musical instrument performance (MIP). These six sounds were uniformly edited, calibrated, and combined using Adobe Audition 2022. All sounds were sampled at 24 bits and 48,000 Hz and their waveforms and spectra are shown in Figure 9. Each sound lasted for 2 min. On one hand, the 24-bit 48,000 Hz sampling can more accurately capture the subtle changes in sounds and present high-fidelity sounds to the participants. On the other hand, the 2 min soundscape experience gives students sufficient time to fully perceive and adapt to each sound stimulus, thereby more accurately evaluating its impact on the restorative experience.

(3) Experimental Procedure

The “restorative theory” mainly includes the stress reduction theory and the Attention Restoration Theory. Therefore, the experimental part of this study mainly considers the impact of sound environment stimuli on human stress and attention (Figure 10).

Experiment 1 (Physiological Measurement): Before the experiment, participants chose a comfortable and relaxed sitting position. We equipped the participants with wearable physiological sensors (Empatica E4) to continuously monitor autonomic responses. Then, the participants sat still for 2 min to test their physiological baseline levels. The experimental steps are as follows: (1) participants perform a 2 min mental arithmetic task to increase stress (stress period). The mental arithmetic task involves addition and subtraction within 500, and participants answer orally. If the participants answer incorrectly, the researchers will re-ask the question until they get the correct answer. The difficulty of the mental arithmetic is adjusted appropriately according to the participants’ responses. (2) We presented a 2 min soundscape stimulus (recovery period) to track physiological recovery and perceive the sound environment without interference. (3) The participants take a one-minute break. Steps 1–3 are repeated 5 times until all six sound environment experiments are completed. The participants take a 10 min break, switch to another set of video scenes, and repeat this step. After the experiment, basic information such as the students’ age and gender is recorded.

Experiment 2 (SART Test): Similar to the above physiological test process, considering that students may become fatigued during long-term experiments, a new group of students was recruited to participate in the experiment. The specific steps are as follows: (1) Participants perform a 2 min mental arithmetic task to increase stress (stress period). The mental arithmetic task involves addition and subtraction within 500, and participants answer orally. If the participants answer incorrectly, the researchers will re-ask the question until they get the correct answer. The difficulty of the mental arithmetic is adjusted appropriately according to the participants’ responses. (2) Participants participate in the SART test. (3) We presented a 2 min soundscape stimulus (recovery period) to track physiological recovery and perceive the sound environment without interference. (4) Participants participate in the SART test again. (5) Participants take a one-minute break and repeat the above steps until all six sound environment experiments are completed. Participants take a 10 min break, switch to another set of video scenes, and repeat this step.

A critical aspect of the experimental design was the control of presentation order. To eliminate potential carryover effects, order effects, and progressive fatigue during the recovery periods, the sequence of the six soundscape stimuli was strictly controlled using a balanced Latin square design. This ensured that each soundscape had an equal probability of appearing at any position in the sequence, and that no specific soundscape consistently preceded or followed another.

3. Results

3.1. The Results of the Field Research

3.1.1. Sound Level Spatiotemporal Characteristics

(1) Temporal Characteristics

Research found that the temporal variation in sound level indicators showed a trend of first rising and then falling (Figure 11). The period from 14:00 to 18:00 was the sound level peak, and 18:00 to 22:00 was the low peak. Among them, L₁₀ − L₉₀ were 5.19, 5.6, 7.52, and 6.39, respectively, with the largest fluctuation difference between 14:00 and 18:00. All indicators gradually increased from 18:00 to 22:00, 6:00 to 10:00, 10:00 to 14:00, and 14:00 to 18:00. The human traffic density in each period might be the main factor determining these changes.

(2) Spatial Characteristics

With the aid of the IDW interpolation method, the spatial distribution pattern of sound level indicators was mapped (Figure 12). From a temporal perspective, 14:00–18:00 and 18:00–22:00 were the periods with the largest ranges of high and low values for each sound level indicator. From a spatial perspective, high sound level indicator areas were concentrated in the planned construction area, leisure and entertainment area, and migratory bird conservation area from 10:00 to 18:00 (Figure 12(B1–B4,C1–C4)). Low sound level indicator areas were distributed in the humanistic landscape area and waterfront recreational area during 6:00–18:00 (Figure 12(A1–A4,B1–B4,C1–C4)). From 18:00 to 22:00, sound levels were generally low, with only the planned construction area and entrance/exit areas maintaining relatively high levels (Figure 12(D1–D4)). The distribution characteristics of L₁₀ − L₉₀ fluctuation differences were generally consistent with L_Aeq and L₁₀.

To address the potential outsized influence of individual measurement points on zone-level comparisons, a leave-one-out sensitivity analysis was conducted. Specifically, Sampling Point 2 in the Sunshine Lawn area was temporarily excluded from the dataset. This point recorded the maximum L₁₀ − L₉₀ values (14:00–18:00) due to temporary localized noise disturbances. Recalculating the zone-level mean sound levels without this point confirmed that the overall spatial distribution patterns and zone-level acoustic differences remained statistically stable. This indicates that the observed acoustic dynamics reflect genuine regional environmental characteristics rather than artifacts of isolated outliers or unbalanced sampling density.

3.1.2. Spatiotemporal Characteristics of Sound Source Harmony

(1) Temporal Characteristics of Sound Source Harmony

Based on the questionnaire survey data (POS, PLS, PFS), the sound harmonicity degree was obtained through processing. After standardization, the daily variation trend of the SHD was obtained (Figure 13). From 6:00 to 10:00, the biophony SHD reached its peak compared to other sound sources and remained at a relatively high level among the five sound source categories across all four time periods; from 18:00 to 22:00, the mechanical sound SHD dropped to its lowest level and remained at a relatively low level among the five sound source categories across all four time periods; the geophony SHD increased over time, reaching its highest level from 18:00 to 22:00, indicating that geophony perception was most prominent during this period. The human activity sound SHD showed a trend of first decreasing and then increasing, which was opposite to the trend of human traffic density in the park; the musical sound SHD first increased and then decreased, which was consistent with the trend of human traffic density.

(2) Spatial Characteristics of Sound Source Harmony

Based on the normalized mean SHD values, a spatial distribution map was generated using IDW (Figure 14). The spatial distribution of harmony varied for different types of sound sources. Biophony and geophony SHDs showed similar spatial distributions, with high values concentrated in the migratory bird conservation area and the humanistic landscape area. Geophony SHD high values were significantly distributed around Aixihu Lake and its secondary water systems. The human activity sound SHD exhibited high values in the leisure and entertainment area and the planned construction area. Mechanical sound SHD low-value areas were mainly distributed in the planned construction area, and high-value areas were mainly distributed in the waterfront recreational area. Musical sound SHD high-value areas appeared in the migratory bird conservation area and the planned construction area.

(3) Global Spatial Autocorrelation

Global spatial autocorrelation analysis was performed on the sound harmonious degree (

Table 3. Global spatial autocorrelation analysis of sound source harmony.

Sound Source Harmony	Moran’s I	z-Score	p
Biophony	0.344	3.603 **	0.000
Geophony	0.255	1.941	0.052
Human Activity Sounds	0.186	3.243 **	0.001
Mechanical Sounds	0.307	2.168 *	0.030
Musical Sounds	−0.075	−0.199	0.842

Note: * z < −1.96 or z > 1.96, ** z < −2.58 or z > 2.58.

). The significance of global autocorrelation is tested using p-values and z-values. When z > 1.96 and z < −1.96 or z > 2.58 and z < −2.58, significant or highly significant spatial autocorrelation for a certain attribute of elements in space is suggested. The results show that, except for musical SHD, which exhibits a dispersed pattern (negative value), other sound source harmonies all show a clustered pattern (positive value). Biophony, human activity sounds, and mechanical sound harmonies all exhibit significant positive spatial autocorrelation. Among them, biophony has the largest z-value (3.603), showing the most significant positive spatial autocorrelation and the largest degree of spatial clustering (Figure 15); human activity sounds and mechanical sounds show significant positive spatial autocorrelation and significant clustering; geophony exhibits a clustered pattern in space, but with a lower degree of clustering.

(4) Local Spatial Autocorrelation

Carry out local spatial autocorrelation analysis on SHD (

Table 4. Local spatial autocorrelation analysis results of sound source harmony.

Sound Source Harmony	Distribution Pattern	Sampling Points
Biophony	HH	6, 7
	LL	11
	HL	10
Geophony	HH	19, 21
	HL	16
	LL	14
Human Activity Sounds	HL	9, 10, 11, 12, 13
	LH	7
Mechanical Sounds	HH	21
	HL	22
	LH	18
Musical Sounds	HL	18
	LH	4

). HH and LL indicate spatial homogeneity of sound source harmony, both exhibiting positive spatial autocorrelation. HL and LH indicate spatial heterogeneity of sound source harmony, both exhibiting negative spatial autocorrelation. In the local spatial model, all sound source perception types show significant spatial autocorrelation. Biophony, geophony, and human activity SHDs have a larger number of sample points with HH and LL clustering patterns; mechanical SHD characteristics mainly show LL clustering patterns; except for human activity sounds, other sound sources exhibit outlier phenomena, with musical sounds being the main category producing outliers.

From the perspective of spatial functional zones, HH clustered sampling points are concentrated in the migratory bird conservation area and the humanistic landscape area. LL clustered sampling points are mainly concentrated in the leisure and entertainment area and the planned construction area. From the perspective of sound source types, sample points where biophony and geophony exhibit LL clustering are mainly in the leisure and entertainment area, while sample points exhibiting HH clustering are mainly in the forest landscape area and humanistic landscape area.

(5) The spatial relationship between SHD and sound level

By conducting a band-wise statistical analysis of the sound level and SHD, the spatial correlation between the two can be obtained (Figure 16). The correlation coefficient Corr ranges from [−1,1]; when Corr is greater than 0, it indicates positive spatial correlation, showing a clustering pattern; when it is less than 0, it suggests negative spatial correlation, showing a dispersed state; when it is equal to 0, it suggests no spatial dependence [46]. The results show that sound level indicators, as well as sound source harmony, primarily exhibit different degrees of negative spatial correlation, with a more remarkable spatial dispersion state. Geophony, human activity sounds, and musical SHD show negative spatial correlation with sound level indicators, with clear dispersion effects, among which the dispersion effect of geophony is smaller and stable.

Biophony harmony suggested a spatially positive correlation with sound level indicators, with the highest correlation coefficient observed for L₉₀ (0.286). This indicates that biophony harmony exhibits a more pronounced clustering tendency in areas with higher L₉₀ values. The relationship between mechanical sounds and sound level indicators is more complex: they show positive correlations with L_Aeq and L₁₀ (0.149, 0.147), while demonstrating significantly negative correlations with L₉₀ and L₁₀ − L₉₀ (−0.426, −0.291).

3.1.3. Soundscape Perceptual Restorative Characteristics

Statistical analysis of the sensitivity characteristics of typical sound source perception in different areas of the wetland park reveals that Figure 17 shows the proportion of people who heard a specific sound source to the total number of samples in that area. Among the six areas of the park, tree canopy bird song had the highest perception level, followed by adult conversations and children playing, and the lowest was singing. Further non-parametric tests were used to examine the differences in the same sound source across different areas, revealing that, except for wind blowing through water plants, other natural sounds showed significant differences in distribution within the park; among human activity sounds, only children playing showed differences; urban traffic sound and water pump sound, both mechanical sounds, showed significant differences across different areas.

To further explore the types of sound sources with restorative benefits in urban parks, the ratio of the frequency with which respondents considered a sound source to have restorative benefits to the frequency with which that sound source was heard was defined as the restorative benefit perception sensitivity of that sound source [38]. The results in Figure 18 show that biophony and geophony have the relatively highest restorative benefit perception sensitivity, followed by musical sounds. The most restorative single sound sources were, in order, tree canopy bird song (0.89), water flowing sounds (0.78), and wind blowing through leaves (0.75). Natural sounds have been widely verified as sound sources with high restorative benefits [47,48,49], consistent with the conclusions of this study. Among different sound source categories, the highest restorative benefit perception sensitivities were for tree canopy bird song in biophony, water flowing sounds in geophony, musical sounds in musical sounds, fitness activity sounds in human activity sounds, and water pump sounds in mechanical sounds.

(1) The relationship between sound pressure level and the evaluation of soundscape perceived restorativeness

This study explored the relationship between the sound pressure level and the mean perceived restorative evaluation of the soundscape at corresponding locations. To control for temporal variance, data from the period with the highest pedestrian density (14:00–18:00) were analyzed. Rather than relying on visual observation, a formal piecewise regression (segmented regression) analysis was conducted to estimate the threshold in the non-linear relationship between L_Aeq and PRSS. The segmented regression model statistically identified a critical breakpoint at 56.4 dB. Prior to this 56.4 dB threshold, the restorative evaluation demonstrated a significant positive correlation with sound pressure level, increasing as L_Aeq increased (slope = 0.153, 95% CI: [0.127, 0.178], R² = 0.845, p < 0.001). Conversely, once L_Aeq exceeded the 56.4 dB breakpoint, the perceived restorativeness exhibited a significant decline with further increases in sound pressure level (slope = −0.039, 95% CI: [−0.049, −0.029], R² = 0.592, p < 0.001). These findings mathematically validate the biphasic impact of acoustic intensity on psychological restoration (Figure 19).

(2) Correlation analysis of sound level, SHD and PRSS

Due to significant differences in the perception of different sound sources, in order to deeply explore the influence of various factors on the perception resilience, we conducted Spearman rank correlation analysis for four sound level indicators, five sound SHDs, and the overall soundscape resilience (i.e., the mean score of PRSS in the third part of the questionnaire) (Figure 20). The results show that biophony and geophony harmony have a significant positive correlation with soundscape perceived restorativeness, while musical SHD shows a positive but not significant correlation. Conversely, all four types of sound pressure level indicators and mechanical SHDs have a significant negative correlation with soundscape perceived restorativeness, with the equivalent sound level L_Aeq having the largest correlation (−0.71). It is worth noting that the L_Aeq shows a strong positive correlation with the L₁₀.

3.2. Experimental Result

We conducted a Kruskal–Wallis test on the baseline physiological indicators (SCL, ST, HR) measured immediately before each of the six soundscape exposures. The results indicated that the p-values of the three physiological indicators were all greater than 0.05. There were no significant differences in the baseline values of each sound environment group under different physiological indicators. This suggests that individuals participating in different sound environments are roughly equivalent at the physiological level (

Table 5. Analysis of differences in baseline values among different scenarios.

Sound Category	SCL	ST	HR
BS	2.15	31.2	72.5
WF	2.08	31.5	71.8
RL	2.22	31.3	73.1
CS	2.19	31.6	70.9
MBC	2.05	31.4	72.2
MIP	2.11	31.5	71.5
F	1.42	1.28	0.96
P	0.925	0.937	0.964

Note: * p < 0.05, ** p < 0.01, *** p < 0.01.

). And, the p-values for all three physiological indicators across the different sound environment baseline periods were greater than 0.05. This statistically confirms that participants successfully returned to a comparable, equivalent physiological baseline before each exposure, validating the 1 min break and our subsequent recovery comparisons.

The Wilcoxon rank sum test indicated (through paired sample analysis of physiological indicators before and after the stress task) that when the pressure was increased, the data of SCL, ST, and SR all rose, suggesting that the mental arithmetic task could induce an increase in stress in middle school students. As shown in

Table 6. Comparative analysis of the benchmark period and the stress period.

Period	SCL	ST	HR
Base	6.25	33.15	71.32
Stress	8.91	32.98	85.47
Z	−5.327	−2.041	−6.184
P	0.000 ***	0.041 *	0.000 ***

Note: * p < 0.05, *** p < 0.01.

, compared with the baseline period of calmness, the participants in the stress period had significantly higher SCL (Z = −5.327, p < 0.001) and HR (Z = −6.184, p < 0.001), while ST significantly decreased (Z = −2.041, p = 0.041). This result consistently indicates that the adopted stress task successfully triggered a significant autonomic nervous stress response, and the stress model was established effectively.

3.2.1. The Restorative Effect of Soundscapes on Physiological Responses

Compared with the stress period, the levels of galvanic skin response and skin temperature decreased significantly in all acoustic scenarios; although the heart rate levels decreased in the recovery period of each acoustic scenario, there were no significant changes (Figure 21).

The generalized estimation equation was used to further explore the changing trend of physiological data during the recovery period. The results showed that, for the main effect of time, there were differences in different time periods in terms of galvanic skin response (Wald χ² = 142.016, p < 0.05), skin temperature (Wald χ² = 153.201, p < 0.05), and heart rate (Wald χ² = 20.833, p < 0.05). The galvanic skin response and heart rate showed a decreasing trend with the increase in time, respectively, being higher than the baseline value by 6.314% and 0.572%. Although the level of skin temperature showed an increasing trend with time, it was lower than the baseline value by 0.186% (Figure 22). There was no interaction effect between the acoustic scene and time on each physiological indicator.

3.2.2. The Restorative Effect of Soundscapes on Attention

The Wilcoxon rank sum test was used to conduct a differential analysis of the physiological indicators of the subjects during the stress period and the recovery period (

Table 7. Analysis of the differences in reaction time and reaction errors before and after soundscape stimulation.

Soundscape	Attention Indicators	Before Stimulation Median (IQR)	Post-Stimulus Median (IQR)	Z	p
BS	response time	276.00 (114.00)	265.00 (122.00)	−6.460 **	0.000
	Reaction error	19.00 (9.00)	16.50 (11.00)	−3.253 **	0.307
WF	response time	268.00 (129.00)	264.00 (121.00)	−1.750 *	0.080
	Reaction error	18.00 (15.83)	17.50 (10.00)	−1.476	0.140
RL	response time	272.00 (117.00)	267.00 (121.00)	−4.168 **	0.000
	Reaction error	18.00 (10.00)	16.00 (12.00)	−0.918	0.358
CS	response time	273.00 (110.00)	264.00 (118.00)	−6.606 **	0.000
	Reaction error	17.00 (11.00)	18.00 (7.00)	−1.021	0.001
MBC	response time	267.00 (123.00)	263.00 (118.00)	−2.958 **	0.003
	Reaction error	18.00 (11.00)	18.00 (10.00)	−1.328	0.026
MIP	response time	274.00 (114.75)	271.00 (116.00)	−4.201 **	0.000
	Reaction error	18.00 (12.00)	17.00 (11.00)	−2.231 *	0.184

Note: * p < 0.05, ** p < 0.01.

). The results of the SART experiment showed that after the subjects were exposed to six types of soundscape perception stimuli, their reaction times significantly decreased, indicating an improvement in reaction speed. This suggests that all six soundscapes have actual attention-restoring effects. After playing the sounds of tree-top birds and musical instruments, the reaction errors of the students significantly decreased.

To compare the effects of different soundscapes on the reaction speed and the recovery of reaction inhibition of the subjects, the change values before and after the soundscape stimulation can be calculated. The larger the change value, the better the recovery effect. The Kruskal–Wallis test showed that there were significant differences in the change values of reaction time among different soundscapes (H = 18.502, p = 0.002). After pairwise comparisons, it was found that the changes in reaction time for BS and CS were significantly greater than those for MBC (Figure 23). Although the change values of reaction errors for different soundscapes were all positive, this change did not have a significant difference (H = 2.028, p = 0.945).

4. Discussion

4.1. Analysis of Field Research Results

4.1.1. Spatiotemporal Characteristics and Urban Green Space Planning

The spatiotemporal dynamics of these sound levels underscore the profound influence of human mobility on urban acoustic ecologies. These acoustic variations align closely with the pedestrian density patterns visualized in our heat maps. Specifically, fluctuations in sound levels across different functional zones strictly parallel diurnal human activity cycles (Figure 8). Consequently, areas characterized by high population density, intense recreational use, and proximity to external traffic corridors consistently generate the most elevated and pervasive acoustic profiles [50].

Spatially, elevated sound levels dominate the planned construction, leisure and entertainment, and migratory bird conservation areas, whereas the cultural landscape and waterfront recreational zones remain distinctly quiet (Figure 12). Specific geographic and environmental variables drive these acoustic extremes. For example, the planned construction area’s immediate proximity to the Aixihu Bridge subjects it to intense urban traffic noise and high pedestrian throughput (Figure 8 and Figure 12). Similarly, peak sound measurements in the leisure area cluster around the Aixihu Tunnel entrance; here, sparse vegetation cover (Figure 4) fails to provide adequate natural acoustic buffering, thereby amplifying local noise. In stark contrast, the cultural landscape and western waterfront areas maintain consistently low sound pressure levels across all time periods. Their scattered wetland island topography and lack of commercial facilities naturally deter heavy pedestrian traffic (Figure 8). Consequently, the park exhibits a clear macro-level acoustic gradient: high-intensity sound environments permeate the northern and southern perimeters (near entrances and traffic corridors), while the central and western sectors successfully preserve low-noise, restorative acoustic conditions.

This highlights a critical challenge in contemporary urban green space planning: the inherent conflict between recreational carrying capacity and acoustic environmental quality. Unlike traditional urban noise control, which relies heavily on static physical barriers, our spatiotemporal mapping suggests that dynamic, time-sharing management strategies are essential. Implementing temporal zoning—such as restricting loud mechanical maintenance or limiting dense crowd activities during critical ecological restoration periods—can effectively mitigate acoustic degradation during peak recreational hours without compromising the park’s public service functions.

4.1.2. Spatiotemporal Characteristics of SHD

The sound harmonious degree (SHD) in Aixihu Wetland Park fluctuates significantly throughout the day (Figure 13), a dynamic largely driven by the functional attributes of each spatial zone. For example, the leisure and planned construction areas demonstrate a stark inverse acoustic relationship: they exhibit severely a degraded SHD for natural sounds (biophony and geophony), while simultaneously recording a peak SHD for anthropogenic noise. This spatial divergence highlights a distinct “acoustic competition effect” [51]. Essentially, competing sound sources actively mask and inhibit one another across the landscape, which fundamentally drives the observed spatiotemporal volatility in soundscape harmony.

The significant spatial clustering of biophony and the spatial dispersion of mechanical sounds reveal complex underlying psychoacoustic mechanisms. Specifically, energetic masking and the phenomenon of “competing sound sources” fundamentally explain the strong negative spatial correlation between SHD and L_Aeq. As demonstrated in cortical processing studies [51], competing auditory streams rely heavily on spatial separation for attentional selection. Within the wetland park, high-intensity mechanical and anthropogenic noise (reflected in L₉₀) physiologically competes with and masks lower-frequency biophony and geophony. Because human auditory processing operates on a limited cognitive bandwidth, the constant cortical effort required to filter out mechanical noise directly degrades the listener’s perceptual sensitivity to—and perceived harmony of—natural sounds. Consequently, the High-High (HH) spatial clustering of biophony harmony observed in the forest landscape is not merely a geographic coincidence; it delineates intact “acoustic habitats” where masking effects are minimized, successfully preserving the soundscape’s ecological and perceptual integrity.

(1) Global Spatial Autocorrelation, Acoustic Connectivity, and Edge Effects

Global spatial autocorrelation analysis reveals that the harmony of different sound sources exhibits varying degrees of spatial clustering and dispersion (Table 5). The significant spatial autocorrelation observed for both biophony and anthropogenic sounds demonstrates that soundscape perception is not randomly distributed; rather, it fundamentally depends on the spatial configuration of the landscape matrix.

Viewed through the lens of landscape ecology, these statistical clusters reveal critical acoustic mechanisms. The strong positive spatial correlation of biophony indicates a high degree of “acoustic connectivity” within the park’s core green spaces. Here, continuous patches of natural vegetation allow biophony to propagate seamlessly, establishing a cohesive and dominant acoustic matrix. In stark contrast, the weaker spatial clustering of mechanical sounds illustrates “acoustic fragmentation.” Boundary “edge effects” primarily drive this fragmentation, as high-intensity urban traffic noise penetrates the park’s perimeters. Consequently, sustaining the global harmony of a wetland soundscape requires meticulously preserving the internal ecological matrix while actively buffering against the invasive penetration of external urban noise.

(2) Local Spatial Autocorrelation, Acoustic Refuges, and Spatial Competition

Local spatial autocorrelation analysis reveals that the distribution of SHD aligns strictly with the park’s functional zoning (Table 6). Consequently, sampling locations with analogous functional attributes consistently exhibit parallel acoustic distribution patterns.

This spatial dependence dictates the formation of distinct “acoustic habitats.” Specifically, the High-High (HH) spatial clustering of biophony and geophony within the forest and cultural landscape zones indicates the preservation of intact “acoustic refuges.” Characterized by dense vegetation and minimal anthropogenic interference, these ecologically buffered areas allow natural sounds to dominate the local acoustic niche free from energetic masking.

Conversely, the Low-Low (LL) clusters of natural SHD in the waterfront recreational and leisure areas highlight severe spatial acoustic competition. In these high-traffic zones, the intensity of human activity sounds and mechanical sounds creates a spatial masking effect that actively suppresses the perceptual clarity of biophony and geophony. This spatial heterogeneity theoretically demonstrates that soundscape harmony is governed not merely by the physical presence of natural sound sources, but by the acoustic carrying capacity of the specific functional zone. Consequently, sustainable green space planning must prioritize the protection of intact HH “acoustic refuges” through the implementation of transitional buffer zones. This strategic zoning is essential to prevent the pervasive masking effects of high-activity areas from degrading the restorative quality of adjacent ecologically sensitive sectors.

(3) Spatial Relationship Between SHD and Sound Level Indicators

Sound level indicators and SHD exhibit varying degrees of negative spatial correlation, characterized by pronounced spatial dispersion (Figure 16). Integrating psychoacoustic energetic masking with the neurophysiological concept of “competing sound sources” fundamentally explains this inverse relationship. As Maddox et al. [51] demonstrated, resolving competing auditory streams relies heavily on spatial separation for attentional selection during cortical processing. At the landscape scale, high-intensity mechanical noise (reflected in elevated LAeq and L₉₀ metrics) physiologically competes with and masks lower-frequency biophony and geophony. Because human auditory processing operates on a limited cognitive bandwidth, the constant cortical effort required to filter out mechanical noise directly degrades a listener’s perceptual sensitivity to—and perceived harmony of—natural sounds. Consequently, these findings establish a critical theoretical link between micro-level neurophysiological processing and macro-level soundscape perception, elucidating exactly how urban traffic noise infiltration actively disrupts the spatial clustering of natural sound harmony.

Among the sound level indicators, L₉₀ exhibits the most significant spatial correlation with various sound harmonious degrees (SHDs). For instance, the SHD of musical sounds demonstrates a strong negative spatial correlation with L₉₀ (r = −0.503). This inverse relationship is primarily due to the masking effect of mechanical noise in environments with high background sound levels, such as planned development and recreational zones. Furthermore, different sound sources respond divergently to sound levels. The SHD of geophony, human activity sounds, and musical sounds all exhibit negative spatial correlations with sound level indicators, accompanied by distinct spatial dispersion effects. Geophony displays a relatively small and stable dispersion effect; this is largely because it is most clearly perceived in secluded areas—such as waterfront leisure zones—that are distant from main roads and entrances, thereby minimizing acoustic masking by other sources. Conversely, biophony SHD demonstrates a positive spatial correlation with sound level indicators, albeit with a weak overall clustering degree. This anomaly is largely attributed to the inherently high amplitude of the unique bird calls within the wetland [52]. Overall, the relationship between sound level indicators and various SHDs is characterized by complex, predominantly negative spatial correlations driven by acoustic masking.

4.1.3. Analysis of the Restorative Characteristics of Soundscapes

(1) Analysis of Restorative Characteristics of Various Soundscapes

The significant disparities in restorative benefits across different sound sources provide compelling empirical evidence for Attention Restoration Theory (ART) [3,47]. Rather than simply restating visitor preferences, these findings reveal a distinct cognitive dichotomy between natural and anthropogenic acoustic environments. The overwhelming restorative dominance of biophony (particularly tree canopy bird song) and geophony (such as flowing water and rustling leaves) is theoretically rooted in their capacity to elicit “soft fascination” [47]. These natural acoustic elements possess complex yet non-threatening informational structures that engage involuntary attention effortlessly, thereby allowing depleted directed attention mechanisms to rest and replenish.

Conversely, the minimal restorative value associated with mechanical and intrusive anthropogenic noises (e.g., urban traffic, park tour vehicles) reflects their cognitively demanding nature. These sounds require active, top-down cognitive suppression, which directly exacerbates mental fatigue. Consequently, urban green space planning must transcend the passive mitigation of noise. Soundscape design should actively cultivate “acoustic refuges” by prioritizing the preservation and strategic introduction of high-restorative soundmarks. Specific spatial interventions—such as designing cascading topographies to introduce continuous geophony or optimizing vertical vegetation structures to create avian habitats—must be recognized as critical public health infrastructure strategies that fundamentally enhance the perceived restorativeness of urban wetlands.

(2) Correlation Analysis of Sound Pressure Level and Restorative Evaluation

The subjective evaluation of soundscapes is also closely related to the average L_Aeq value. For instance, the critical sound pressure level corresponding to the acoustic comfort threshold that people experience in open public spaces in European cities is 57 dB [53]. The corresponding sound pressure level in typical urban green spaces in Chengdu is 77 dB [54], and the sound pressure level in Han Chinese Buddhist temples is 60 dB [55]. Most existing studies have focused on establishing the relationship between the comfort level of soundscapes and various environments. As a subjective assessment, the perception of restorative soundscapes should also have a correlation with L_Aeq. This study constructs the relationship between the physical environment and subjective perception, and supplements Guo scholar’s research [38] on the role of soundscapes in alleviating tourists’ stress in urban parks. The results show that there is a positive correlation between L_Aeq and the PRSS before 56.4 dB. Beyond this threshold, the perception of restorative soundscapes decreases as L_Aeq increases.

While the segmented regression mathematically identified a 56.4 dB L_Aeq threshold for restorative benefits, this value must be interpreted as context-dependent and exploratory. This specific threshold reflects the acoustic tolerance within a Chinese urban wetland park during peak recreational hours. Acceptable noise limits for psychological restoration are likely to vary significantly across different ecological settings (e.g., dense forests vs. open waterfronts), park typologies (e.g., historical gardens vs. urban squares), and cultural contexts with differing baseline urban noise exposures. Future cross-regional and cross-cultural studies are necessary to validate the environmental universality of this specific decibel threshold.

(3) Correlation analysis of sound level, SHD and PRSS

Viewed through the lens of Attention Restoration Theory (ART) [3,47], these findings provide robust empirical evidence for how acoustic environments modulate cognitive fatigue. The equivalent sound level L_Aeq emerged as the strongest negative influencing factor (−0.71) on restorative benefits. Theoretically, high noise levels—predominantly anthropogenic and mechanical—require directed attention to be effectively suppressed, which directly degrades the “Being Away” and “Fascination” components of the restorative experience [47]. Conversely, biophony (especially bird song) and geophony exhibit strong restorative potential because their complex, non-threatening acoustic structures effortlessly engage “soft fascination,” allowing directed attention mechanisms to rest.

Furthermore, the identification of the 56.4 dB L_Aeq threshold is a critical novel contribution to urban green space planning. It provides a quantitative physiological boundary where the restorative benefits of natural sounds are overwhelmed by the stress-inducing properties of anthropogenic noise. While previous studies have established 57 dB as a general acoustic comfort threshold in European public spaces [53], our findings extend this threshold specifically to psychological restoration in Chinese wetlands. This suggests a cross-cultural convergence in human acoustic tolerance. It strongly argues that urban green infrastructure strategies must transition from merely introducing natural sounds (e.g., attracting birds) to actively mitigating background mechanical sounds (controlling L₁₀) to successfully preserve the acoustic restorative capacities of these environments.

These findings critically align with and expand upon recent global syntheses in soundscape ecology. For instance, recent high-impact reviews of international protected areas have quantitatively established that natural biophony not only mitigates perceived noise annoyance but fundamentally alters health outcomes by lowering cardiovascular stress markers [56]. While much of the existing literature heavily focuses on North American or European contexts, our empirical data from a rapidly developing Chinese urban wetland provides crucial cross-cultural validation. It demonstrates that the acoustic carrying capacity and the concept of “acoustic refuges” are universal imperatives in global urban green space planning.

4.2. Analysis of Experimental Results

4.2.1. Physiological Measurement Experiment

From a physiological perspective, the rapid recovery from acute stress facilitated by the urban wetland soundscapes extends beyond simple auditory preference; it reflects a deep-rooted evolutionary mechanism. The pronounced reduction in skin conductance levels (SCL) induced by biophony (e.g., tree canopy bird song) and geophony (e.g., rustling leaves) is driven by the autonomic nervous system’s response to environmental acoustic cues. According to biophilia hypotheses, the complex, harmonic, and non-threatening frequencies inherent in these natural sounds signal ecological safety [47,57]. This acoustic input effectively inhibits sympathetic nervous system arousal (the “fight-or-flight” response) and catalyzes a rapid shift toward parasympathetic dominance, thereby accelerating physiological recovery [19,24].

Further exploration of the changing trends of physiological indicators during the recovery period reveals that the duration and changes in soundscapes represent the nervous system’s adaptation to the soundscapes. The finding that peak recovery efficiency for physiological indicators (SCL and HR) occurs within the initial 60 s of soundscape stimulation provides novel temporal granularity to soundscape research, which traditionally aggregates pre- and post-exposure effects. This rapid response aligns perfectly with Ulrich’s Stress Recovery Theory (SRT) [58], which posits that unthreatening natural environments elicit immediate, unconscious autonomic nervous system responses. Our controlled tracking demonstrates that natural soundscapes act as rapid physiological catalysts, swiftly shifting the autonomic balance from sympathetic arousal (stress) back to parasympathetic dominance (relaxation). This implies that even transient, short-term auditory exposures to high-quality natural soundscapes in high-density urban settings can yield significant physiological public health dividends.

Furthermore, the divergent temporal trajectories of these physiological indicators highlight distinct layers of autonomic regulation. For instance, the immediate and steep decline in heart rate at the onset of natural sound stimulation signifies a rapid “vagal brake” application—a primary parasympathetic reflex triggered by acoustic safety cues. This temporal sensitivity underscores that different physiological pathways adapt to restorative acoustic stimuli at varying speeds, providing a more granular understanding of human–soundscape interactions.

4.2.2. Analysis of SART Test Experiment

The cognitive improvements observed in the SART experiment reveal the specific neurocognitive pathways through which different soundscapes intervene in attention regulation. The universal reduction in response times across all six soundscapes suggests that acoustic isolation from acute stressors generally lowers cognitive load, facilitating basic processing speed. However, the reduction in reaction errors—a direct measure of executive inhibitory control—was exclusively triggered by bird song and musical performances.

Mechanistically, this highlights two distinct restorative pathways for cognitive inhibition. Bird song delivers “soft fascination”, engaging involuntary attention through rich, non-taxing acoustic variations [47]. This effortless engagement permits the depleted prefrontal cortex to rest and replenish its executive resources, leading to fewer impulsive errors. Conversely, musical performances likely enhance inhibitory control through rhythmic predictability and structural coherence. This acoustic structure provides a cognitive “scaffold” that sustains directed attention without inducing mental fatigue. These findings theoretically demonstrate that while any pleasant sound environment may speed up basic reactions, highly structured or evolutionarily familiar acoustic signals are fundamentally required to restore complex executive inhibition.

From a psychoacoustic perspective, this cognitive enhancement is increasingly supported by the recent international neuroimaging literature. Recent studies indicate that exposure to high-fidelity natural soundscapes can modulate functional connectivity within the brain’s default mode network, which is responsible for rumination and mental fatigue [59]. Our physiological and cognitive (SART) findings physically manifest this neural shift at the behavioral level. This expands the traditional Attention Restoration Theory (ART) by proving that complex, non-threatening acoustic signals act as a proactive cognitive scaffold, rather than merely a passive background, effectively isolating the central nervous system from the pervasive energetic masking of urban mechanical noise [60].

4.3. Innovation and Limitations

This study makes four core innovations: (1) an integrated framework combining spatiotemporal acoustic mapping, sound source harmony spatial analysis, and multi-dimensional restorative effect experiments; (2) verification of the boundary conditions and dual physiological–cognitive restorative mechanism of bird song in real urban environments, advancing beyond laboratory-based findings; (3) a statistically validated 56.4 dB L_Aeq threshold for soundscape restorative benefits, with translatable strategies for urban sustainable planning; and (4) under the stimulation of the sound environment, the recovery effect of the subjects was the best within 60 s.

This study presents four main limitations. First, the single-site focus and exclusion of multi-sensory synergies limit broader generalizability. Second, the 100 m sampling grid resulted in unbalanced sample sizes across unequally sized functional zones, necessitating stratified random sampling in future designs. Third, data collection was restricted to optimal-weather holidays, capturing isolated peak-use cross-sections rather than continuous spatiotemporal characteristics, thereby excluding weekdays and winter conditions. Finally, regarding the experimental design, it must be acknowledged that the laboratory experiments exclusively recruited university students (mean age ~24 years). As indicated by our field questionnaire data, this cohort does not fully capture the diverse demographic profile of actual park visitors (who encompass a wide range of ages, occupations, and stress histories). Consequently, physiological and cognitive responses to soundscapes may vary among broader populations, which slightly limits the absolute generalizability of the experimental findings.

Future research will conduct multi-case cross-regional comparative studies to verify the generalizability of our core conclusions and explore the multi-sensory synergistic mechanism of restorative urban environments to support healthy and sustainable city development.

5. Conclusions

This study investigated the spatiotemporal characteristics and restorative benefits of urban wetland soundscapes. Using a mixed-methods approach of field surveys and laboratory experiments, we systematically analyzed the spatial distribution of sound levels and perceived sound source harmony and quantified their physiological and cognitive restorative capacities. The main findings are as follows:

(1) Human mobility primarily drives the significant spatiotemporal variations in both sound level indicators and the SHD across the wetland park.

(2) Biophony, geophony, and anthropogenic sounds exhibit strong spatial competition; the SHD shows significant spatial autocorrelation in both global and local models, while correlating negatively with sound levels.

(3) Restorative potential varies significantly among sound sources, with biophony (specifically bird song) demonstrating the highest efficacy for psychological and physiological recovery.

(4) PRSS correlates positively with L_Aeq up to a threshold of 56.4 dB, beyond which it declines. The perceptual harmony of natural sounds strongly dictates this subjective evaluation.

(5) Physiologically, short-term exposure to natural soundscapes (e.g., bird song, rustling leaves) rapidly alleviates stress. This restorative effect peaks within the initial 60 s of exposure, evidenced by significant decreases in skin conductance level (SCL) and heart rate (HR).

(6) Cognitively, natural soundscape stimulation shortens SART response times, indicating improved reaction speeds. Furthermore, specific acoustic elements like bird song and musical performances reduce reaction errors, demonstrating enhanced sustained attention and inhibitory control.

Beyond advancing soundscape ecology, these findings offer actionable, data-driven implications for sustainable urban green space planning. Rather than relying on broad strategic concepts, urban planners should implement the following targeted interventions directly linked to our empirical outputs:

(1) Quantitative Acoustic Zoning Based on the 56.4 dB Threshold: The identified 56.4 dB LAeq threshold must serve as a strict regulatory benchmark. Planners should establish “acoustic buffer zones” ensuring that background mechanical noise in restorative landscapes is actively mitigated below this 56.4 dB limit, thereby maximizing the Perceived Restorativeness Soundscape Scale (PRSS) and unmasking natural biophony.

(2) Spatial–-Temporal Trail Design Based on the 60-Second Recovery Peak: Because physiological recovery (evidenced by SCL and HR reductions) peaks within the initial 60 s of exposure to natural sounds, park trail networks should be spatially calibrated. Planners should design paths that guarantee visitors experience uninterrupted natural soundscapes for at least 1 min intervals (equivalent to approximately 60–80 m of walking distance) between high-traffic nodes.

(3) Targeted Source Management Driven by Spatial Correlation (r = −0.503): Given the strong negative spatial correlation (r = −0.503) between L₉₀ and musical sound harmony, spatial planning must prevent energetic masking. Planners should physically separate mechanical infrastructure (e.g., park vehicles, maintenance equipment) from designated cultural spaces to protect the High-High (HH) spatial clustering of high-value natural and humanistic sounds.