Introduction

In the twenty-first century, urbanization has accelerated dramatically, bringing unprecedented challenges for public health. Today, more than half of the global population lives in cities, and this proportion is projected to reach about 70% by 2050 (Nieuwenhuijsen, 2024). Dense urban living environments often heighten stress and exposure to environmental hazards, contributing to a rise in chronic health issues. Notably, city dwellers face significantly higher risks of mental health disorders compared to rural residents—up to a 21% greater prevalence of anxiety and a 39% higher prevalence of mood disorders, along with roughly double the risk of schizophrenia (Gruebner et al., 2017). Addressing the health and well-being of growing urban populations has thus become a critical priority worldwide.

Urban green spaces such as parks are increasingly recognized as essential infrastructure for healthier cities. A broad body of evidence indicates that access to parks can improve both physical and mental well-being (Zhang et al., 2022). For instance, the World Health Organization has noted that urban parks “can promote mental and physical health, and reduce morbidity and mortality” by providing opportunities for psychological relaxation, stress alleviation, physical activity, and social cohesion (Zhao et al., 2024). In response, city planners and policymakers are seeking ways to integrate more green space into dense urban fabrics. In China, for example, the city of Chengdu has launched an ambitious “Park City” initiative to build a “city within a park,” weaving extensive greenways and ecological spaces throughout the metropolis to improve residents’ quality of life (Wan et al., 2024). This approach reflects a growing consensus that well-designed urban parks can serve as preventive public health infrastructure, alleviating some of the mental and physical strains of urban life.

Despite widespread recognition of green space benefits, critical knowledge gaps remain in understanding how urban parks can be optimized to maximize health outcomes. Most prior studies offer broad correlations—for example, linking the amount of nearby green space to general health indicators—while overlooking the nuanced human experiences that actually drive these benefits (Li et al., 2024). In particular, the multisensory qualities of park environments have been largely neglected. Research to date has focused predominantly on visible landscape features, paying far less attention to auditory, olfactory, and tactile stimuli that also shape a visitor’s experience (Kou et al., 2025). This leaves an incomplete picture of restorative environmental design: for instance, how the sound of birds or the scent of vegetation might reduce stress, or conversely how traffic noise could undermine an otherwise green setting. Moreover, user behavior is seldom integrated into these analyses—people experience parks not just through their senses but through their actions (e.g. walking, socializing, exercising), yet many studies fail to capture what visitors actually do in parks and where they do it (Tavakoli et al., 2025). The lack of high-resolution data linking specific park features, sensory experiences, and user behaviors to concrete health outcomes represents a significant research gap and a barrier to evidence-based urban park design.In this study, we address these gaps by introducing a Public Participation GIS (PPGIS)-embedded mixed-methods framework to decode the “sensory-behavioral pathway” linking urban park experiences with public health outcomes. Our approach integrates spatial mapping, quantitative surveys, and qualitative insights in an innovative way, using data from urban parks in Chengdu, China. By embedding a PPGIS protocol into the research design, we capture fine-grained information about where and how people interact with park spaces, which aspects of the environment engage their senses, and how these interactions relate to self-reported health and well-being measures. This design allows for a rigorously detailed analysis of the park experience that goes beyond traditional metrics.

Literature review

Parks and health: from historical observations to modern theories

The nexus between urban parks and population health has been acknowledged since the Middle Ages, marked by the recognition of gardens’ therapeutic and rehabilitative qualities (Mooney et al., 2009). Historical events, notably plagues and mass migrations, posed significant challenges to the development of these healing gardens. The advent of Romanticism, with its reemphasis on the natural environment’s role in physical and mental recuperation, heralded the resurgence of healing gardens in Europe (Lei et al., 2011). The Industrial Revolution in the late 18th century, characterized by urban environmental degradation, further amplified public health concerns. This period was pivotal in solidifying the integral relationship between urban parks and public health, reflecting a transition from anecdotal observations to more structured theories and practices (Zhao et al., 2021).

Frederick Law Olmsted, a forerunner in the urban park beautification movement, significantly influenced subsequent generations with his efforts in enhancing population health through urban park development. His work transcended mere aesthetic improvements, highlighting the pivotal role of green spaces in public health enhancement (Liu et al., 2021). The mid-20th century witnessed a broader acknowledgment of urban parks’ health benefits (Chen et al., 2024), aligning with the World Health Organization’s expanded definition of health as a state of complete physical, mental, and social well-being (Jin et al., 2022; Tzoulas et al., 2007).

The late 20th century marked a further deepening in our understanding of the natural environment-health nexus through Ulrich’s model of psychological evolution and Kaplan’s attentional restoration theory. Ulrich demonstrated that natural environments substantially reduce stress, thus promoting health (Ulrich, 1984). Kaplan’s theory, emphasizing “restorative environments” for mental fatigue recovery, identified natural settings as typically restorative (Kaplan, 1995). Subsequent research has consistently supported these theories, underlining the positive effects of natural environments on population health (Wang et al., 2024).

Despite these theoretical advancements, there remains a research gap concerning the specific mechanisms through which park environments influence health recovery, particularly regarding how multi-sensory experiences and behavioral preferences affect this process. This gap highlights the need for further exploration into the intricate interactions between spatial multisensory perceptions, behavioral preferences, and health outcomes in park environments.

Multi-sensory experiences and behavioral preferences: a new lens on urban park health effects

The health benefits of parks are intricately linked to how populations perceive and interact with these environments. Understanding the environment is a multi-stage process involving information acquisition, cognition, and representation, where initial spatial perception is influenced by the sensory responses—visual, auditory, tactile, olfactory, and gustatory (Li et al., 2023). While the brain organizes and processes sensory data, creating a comprehensive experiential and emotional connection with the surroundings, current research primarily focuses on unidimensional sensory studies based on either visual or auditory landscapes (Zhu et al., 2021), while comprehensive research on multisensory experiences remains lacking (Xi et al., 2020). This leaves a significant gap in our understanding of the multi-sensory landscape.

Human sensory experiences, encompassing sight, hearing, smell, taste, and touch, are interdependent and collectively crucial for health promotion. Multi-sensory stimulation in natural settings has been empirically shown to alleviate stress, improve mood, and aid in psychological recovery. However, specific strategies for harnessing these sensory experiences to enhance health in urban park environments remain underexplored (Ziegler, 2015). Furthermore, the behavioral preferences of individuals reflect not just multi-sensory experiences but also the interaction of physiological and psychological factors with their environment (Yakinlar et al., 2022). This interplay is especially evident in urban parks, which not only showcase how individuals engage with their environment but also how this environment contributes to health recovery. Despite its importance, there is a notable gap in research on effectively integrating these behavioral preferences into urban park design to maximize health benefits. This lack of understanding underscores a critical area for further investigation to optimize the health-promoting functions of urban parks.

Limitations of traditional research methods and prospects of new technologies

Traditional research methods in studying the spatial environment and health effects in parks, primarily relying on observational, experimental, and questionnaire techniques, face significant constraints (Ye et al., 2022). These methods often suffer from data inadequacies, low accuracy in ratings, and limited spatial resolution, which hinder a comprehensive understanding of complex spatial dynamics (Ives et al., 2017). The inherent limitation is their inability to capture real-time, crowd-site responses effectively, thus failing to reflect the nuanced spatial interactions in park environments (Schirpke et al., 2016). Moreover, the reliance on researcher’s subjective perceptions in designing questionnaires and experiments introduces potential biases, compromising the validity of the findings (Zhao et al., 2020). While these traditional methods have laid the groundwork for understanding park spaces and health, their limitations underscore the necessity for more dynamic, innovative, and technologically-driven research approaches that can capture the complexities of human-environment interactions more accurately.

The advent of mobile terminals and web mapping technologies has significantly enhanced the potential of smartphones in collecting spatial pattern data related to crowd behaviors (Clemente et al., 2019; Korpilo et al., 2018). These technologies facilitate the mass sharing of public geographic data, thereby bolstering public engagement in Geographic Information System (GIS) development and providing novel data sources for urban planning and decision-making (Peng et al., 2019). However, while this smartphone-based approach has seen extensive application in North America, Europe, and Australia, its adoption in urban research, particularly in developing countries like China, remains limited (Dong et al., 2014). Employing smartphone technology and Public Participation Geographic Information System (PPGIS) methods in studying the impact of park environments on health not only addresses the shortcomings of traditional research methods but also introduces fresh perspectives and tools for urban research. This shift towards advanced technological methods promises more nuanced and comprehensive insights into the interaction between urban spaces and public health.

This study seeks to transcend traditional research methods in exploring the environmental and health effects of urban parks by adopting a novel multi-sensory and behavioral perspective. While previous research has primarily concentrated on single-sensory experiences, our approach delves into the complex interplay of multi-sensory experiences and behavioral preferences. Traditional studies often fall short in capturing the dynamic nature of crowd behavior in real-time urban park environments. To address these gaps, our study, centered around nine typical urban parks in Chengdu City, utilizes cutting-edge methodologies including Public Participation Geographic Information System (PPGIS), latent category analysis, and structural equation modeling.

Specifically, we implemented an interactive PPGIS survey in which park visitors pinpointed on a digital map the locations where they engaged in various activities and described the associated sensory impressions (visual, auditory, olfactory, etc.) at each spot. Each mapped point is linked to the participant’s reported emotional or health response, giving rich geographic context to subjective experiences—in other words, people’s qualitative perceptions acquire precise spatial coordinates. This high-resolution mapping approach goes further than conventional surveys that might simply ask how often someone visits a park; instead, it captures where and what people actually experience, in real time and space. We complemented the PPGIS data with structured questionnaires and interviews about participants’ health status and well-being (e.g. perceived stress reduction, mood improvement), yielding a robust mixed-methods dataset. Crucially, our sample included a diverse cross-section of residents, allowing us to examine how sensory and behavioral patterns—and their health impacts—differ across demographic groups. Such stratification yields insights into whether, for example, older adults benefit differently from certain park features than younger visitors, ensuring that our findings inform inclusive and evidence-based design strategies.

By integrating these elements, our study provides new insights that are both theoretical and practical. The key contributions of this work are as follows:

  1. I.

    Methodological innovation: We develop a novel PPGIS-based mixed-methods protocol for capturing multisensory and behavioral data in urban parks at high spatial resolution, demonstrating an effective way to study human–environment interactions that traditional surveys or GIS analyses alone cannot achieve.

  2. II.

    Empirical insights: Using this approach, we empirically unravel how specific sensory experiences (sights, sounds, smells, etc.) and visitor behaviors interact to influence health and well-being outcomes. This evidence illuminates the proposed sensory-behavioral pathway, clarifying why and how certain park features contribute to stress reduction and other health benefits.

  3. III.

    Inclusivity and differentiation: Through demographic stratification of the data, we reveal important differences in how various groups (e.g. different ages and genders) perceive and benefit from urban parks. These findings highlight the need for inclusive park planning that considers diverse sensory preferences and usage patterns in order to equitably promote public health.

  4. IV.

    Practical guidance: We translate our findings into actionable recommendations for urban planners and policymakers. By identifying which sensory stimuli and spatial features are most conducive to positive health outcomes, we provide an evidence-based framework for designing and managing urban parks to maximize their health-promoting potential.

Methodology

Study area

Chengdu is located in the western part of the Sichuan Basin in southwest China and the eastern edge of the Qinghai-Tibet Plateau(30°05′-31°26′N, 102°54′-104°53′E)(Fig. 1).

Fig. 1
figure 1

Distribution of the study area and park locations.

Full size image

This study selected nine urban parks in Chengdu as case studies (Table 1). These parks are situated in key urban districts, including Qingyang District, Jinniu District, Chenghua District, and Jinjiang District, etc., with areas ranging from 4.71 hm2−30.89 hm2. The selection encompasses a variety of park types, including comprehensive parks, classical garden parks, ecological wetland parks and other types. This diverse selection reflects the overall state of parks in the main urban area, considering factors such as distribution area, sample size, scale and type.

Table 1 Information and characteristics of urban parks.
Full size table

Experimental design

Theoretical models

The creation of a health-promoting environment in urban parks is an intricate process, shaped by the interplay of people, the environment, and behavior. These elements form a symbiotic system, addressing the need for health restoration (Huang et al., 2022). This study focuses on the urban park environment’s impact mechanism on health recovery. As per our literature review, spatial sensory perceptions—sight, hearing, smell, taste, and touch—play a crucial role in facilitating psychological, physiological, and social recuperation (Tan et al., 2021). These dimensions are interrelated, together constituting the unique characteristics of the restorative environmental impact mechanism in urban parks.

However, the relationship between urban parks’ restorative environments and health recovery is not straightforward but influenced by various mediating variables. One key mediator is the public’s engagement in diverse activities within parks (Limin et al., 2016). Therefore, in this study, we first define the restorative environment as a spatial environment that promotes individual physiological and psychological health recovery through multisensory stimulation (including visual, auditory, olfactory, and tactile) and behavioral participation, in which the behavioral preferences of park visitors are taken as mediating variables.

Based on these insights, we developed a theoretical model (Fig. 2) to investigate the underlying principles driving the restorative impact of urban parks. The research hypotheses proposed are:

Fig. 2
figure 2

Research theoretical model and research hypothesis.

Full size image

H1: The perceived spatial environment of urban parks significantly affects population health recovery.

H2: The perceived spatial environment of urban parks significantly influences crowd behavioral preferences.

H3: There is a significant correlation between population behavioral preferences and health recovery.

H4: Population behavioral preferences play a mediating role in the impact of the perceived spatial environment of urban parks on health recovery.

Questionnaire design

The questionnaire survey was conducted over an 8-month period, from July 2022 to March 2023, in various urban parks. During this period, visitors willing to participate were randomly selected from urban parks, and after signing a written informed consent form, they were surveyed, resulting in a collection of 598 valid responses.

The questionnaire was structured into four distinct sections:

Demographics and Health Status: This section gathered basic information about the respondents, including gender, age, occupation, education level, duration of residency, as well as psychological and physical states. The psychological state was assessed using the Warwick-Edinburgh Positive Mental Health Scale (WEM-WBS) with a 5-point scale for evaluation (1 = never to 5 = always). For physical state, we utilized the ‘General Health Status’ section from the SF-36 Health Survey Short Form.

Spatial Perceptions: Respondents were asked to rate their perceptions of various auditory, visual, olfactory, and tactile factors in the park environment (Table 2). Ratings were on a 3-point scale ranging from 1 (uncomfortable) to 3 (comfortable).

Table 2 Selection of indicators.
Full size table

Behavioral Preferences: This section focused on capturing the types of activities respondents engaged in, their frequency of park visits, modes of transportation to the park, and duration of their stay.

Health Recovery Assessments: The final part of the questionnaire sought to evaluate the visitors’ mental and physical health recovery experiences in the park.

This comprehensive questionnaire design aimed to gather a holistic understanding of the visitors’ experiences, perceptions, and health outcomes in urban parks, providing a robust dataset for further analysis.

Implementation of the PPGIS study

This study employed the PPGIS method to precisely map visitors’ sensory experiences to spatial locations, thereby overcoming the limitations of traditional questionnaires. Through an interactive online map, participants could mark specific locations within the park where they experienced particular sensations or engaged in activities in real time. This approach not only significantly enhanced spatial data accuracy but also authentically captured how individuals achieved physical and mental recovery within diverse environmental elements. It provides high-resolution, geolocated empirical evidence for analyzing the spatial mechanisms linking sensory experiences, behaviors, and health outcomes.

This study conducted PPGIS from April to July 2023 to collect on-site sensory experience and spatial behavior data from visitors in urban parks. Given Chengdu People’s Park’s strongest representativeness and comprehensiveness among the nine selected urban parks, and considering the high resource demands of PPGIS fieldwork, this park was ultimately designated as the sole implementation site (Fig. 3).

Fig. 3
figure 3

Chengdu people’s park.

Full size image

Building upon questionnaire surveys, the study recruited 60 volunteers through stratified sampling for field research. Participants were evenly divided into two age groups: youth (under 35 years old) and middle-aged/elderly (35 years old and above), with balanced gender ratios to ensure demographic diversity. All participants signed written informed consent forms prior to the study, clearly acknowledging the research objectives, data usage, and their rights.

Prior to fieldwork, the research team conducted group training sessions to familiarize volunteers with the survey procedures, objectives, and the data collection tool PinSurvey software. To prevent interference among participants, each volunteer was required to independently complete spatial marking and data uploading. Participants were encouraged to document as many different types of spaces as possible to ensure spatial coverage and representativeness of the data.

We employed PinSurvey to collect user activity and questionnaire data. PinSurvey is a mobile mini-program specifically designed for participatory geographic information collection. It automatically records GPS coordinates while taking photos and displays customized questionnaires, enabling quick data entry through a “one tap, one photo, one answer” process. PinSurvey features a launch screen and a map interface where participants can freely place spatial points. Participants were instructed to explore the park independently, photographing spatial environments. The software automatically recorded GPS coordinates at each photo location and guided participants through three categories of online questions: (1) perceptions of the space; (2) potential activities; (3) overall impressions. Geolocation data was solely used to analyze visitor perceptions and behavioral patterns across different spaces, without involving trajectory tracking. De-identification was performed during data processing to safeguard privacy.

Data analysis

Potential categories model with multiple logistic regression models

To investigate the variations in spatial perceptions, behavioral preferences, and health effects among individuals in urban parks, this study employed Mplus to perform latent category analysis. The analysis enables the explanation of associations between exogenous categorical variables by employing a minimal number of categories, while ensuring local independence between exogenous variables. The main fit metrics used for mode include Akaike Information Criterion (AIC), Bayesian Information Criterion (Yang et al.), Adjusted BIC (aBIC), Entropy, Lo-Mendell-Rubin (LMR), and Bootstrapped Likelihood Ratio Test (BLRT) (An et al., 2023). Smaller values of AIC, BIC, and aBIC indicate better model fit. Entropy, ranging from 0 to 1, is a crucial indicator for evaluating the accuracy of category delineation. LMR and BLRT are employed to compare the model fitting of n-1 and n categories. If the p-value corresponding to these indicators reaches a significant level, it suggests that the model with n categories is superior to the model with n-1 categories. Furthermore, to examine the influence of demographic variables on the variations in individuals’ perceptions, multivariate logistic regression analysis was conducted using SPSS 26.0. Eight population social attribute indicators, including gender, age, and education, were considered as independent variables, while each potential category served as the dependent variable.

Structural equation modeling

Structural equation modeling (SEM) is a powerful tool that examines the relationship between independent and dependent variables, while accounting for measurement error in both variables (Buchel et al., 2015). This approach is particularly advantageous for analyzing complex relationships and the degree of influence. In this study, we utilized AMOS to construct a model that explores the mechanism of how urban parks influence health. Building upon the previous theoretical model, we identified three key variables: the independent variable ‘spatial perception’, the mediator variable ‘behavioral preference’, and the dependent variable ‘health effect’ (Fig. 4). The spatial perception variable comprised four measures: auditory, olfactory, tactile, and visual factors. The measurement model for behavioral preferences consisted of four measures: frequency of use, mode of travel, type of activity, and length of stay. The health effect measurement model primarily included two measures: physiological effects and psychological effects.

Fig. 4
figure 4

Pre-conditioned model for structural equations.

Full size image

To estimate and assess the model parameters, we selected several indicators, such as Chi-Square (χ2), Degree of Freedom (df), Root-Mean-Square Error of Approximation (RMSEA), Incremental Fit Index (IFI), Tucker-Lewis Index (TLI), Comparative Fit Index (CFI), Goodness of Fit Index (GFI), and Adjusted Goodness of Fit Index (AGFI) (Wen et al., 2004). If the model does not meet the evaluation criteria, we will make necessary revisions and reevaluate it until the desired results are achieved. Once the model passes the evaluation, we will test the hypotheses for validity and provide a detailed explanation of the model’s results.

Spatial analysis and identification of landscape elements

The kernel density analysis tool in ArcGIS is used to calculate the density of elements in their surrounding neighborhoods, creating a continuous and smooth surface based on the original discrete distribution of spatial points. This enables the visualization of the extent and concentration of discrete point distributions. In this study, spatial point data collected from the PPGIS research were entered into ArcGIS 10.5, and the kernel density analysis tool under the Spatial Analyses tool was selected to process the data. As a result, various types of spatial distribution maps were obtained to reveal the characteristics of the differences in the spatial distribution of People’s Park in a more comprehensive and intuitive manner.

To further explore the relationship between spatial perception, behavioral activities, and landscape elements, this study first annotated the photos online using the EasyDl platform. When a photo had a certain landscape type, it was recorded as 1, and vice versa, it was recorded as 0. The frequency of occurrence of each type of landscape element was also summarized. The landscape categorization criteria were mainly based on Zhang’s categorization methods (Zhang et al., 2022). his classification method was chosen due to its suitability for small and medium-sized parks, providing more accurate results based on the composition and spatial distribution characteristics of landscape elements. Zhang’s paper demonstrated that the MaxEnt model results had an Area Under the Curve value of 0.9 or higher, indicating a high classification accuracy for these landscape elements. Taking into account the actual situation of People’s Park, the landscape types were categorized into six major categories and twenty subcategories (Table 3).

Table 3 Landscape type indicators.
Full size table

Subsequently, this study utilized SPSS 26.0 to conduct correspondence analysis. Correspondence analysis allows for the examination of the relationship between row and column variables by grouping them and reducing dimensionality. In this study, the researchers first coded spatial perception, behavioral activities, and landscape elements. Then, the spatial perception evaluation and behavioral activity variables were assigned to the row checkboxes and column checkboxes, respectively. The values for the row variables ranged from 1−4, representing the four types of spatial perception, while the column variables represented the 20 subcategories of landscape elements, with values ranging from 1−20. The association pattern diagram of the variables was obtained through subsequent operations. The plot results from the origin (0, 0), and the position of the scatter on the plot indicates the degree of association between the row and column categories. When the scatters of row and column categories are close, far from the origin, and in the same orientation or region, it suggests a strong positive association between the row and column categories. Conversely, if the scatters are far from the origin, and in opposite orientations or regions, it indicates a strong negative association between the row and column categories.

Results

Characterization of the social attributes of the population

Among the respondents (Table 4), there were slightly more females (50.7%) than males (49.3%). In terms of age structure, the majority of respondents were aged 18−25 (36.5%), followed by those aged 26−35 (23.9%). The number of older respondents aged 60 or above was relatively low (10%), but still provided representative data. The respondents exhibited a high level of education, with 86.8% having completed high school or above. They represented a diverse range of occupations, including civil servants, corporate employees, freelancers, and full-time students. In terms of residency, the majority of respondents were residents of Chengdu (81.3%), with a significant proportion having lived there for >3 years (63.3%). Furthermore, the vast majority of respondents reported being in good physical and psychological condition, accounting for 91.0% and 87.8% respectively.

Table 4 Statistical table of demographic characteristics.
Full size table

Analysis of population differences

In order to examine the significant explanatory power of latent category variables for the various sensory cognitive, behavioral, and health restoration effects, this study employed latent category analysis. This approach takes into account the distinct distribution patterns of these variables.

For spatial perception, the study estimated the model parameters for the number of potential categories 1–5. Comparing the fitness test metrics of the models with different numbers of potential categories, it was found that the 4-category model provided more accurate delineation. This model exhibited relatively small Log(L), AIC, BIC, ABIC, and Entropy values above 0.9. Additionally, the LMR and BLRT values were significant at the p < 0.001 level, while the LMR value for the 5-category model was no longer significant. Therefore, the 4-category model was ultimately chosen, and response probability plots for the four potential categories on the 48 options were obtained (Fig. 5a). The scores for each option in category C1 are at a moderate level, and the trend of each dimension was similar to that of C4, but relatively speaking, the scores of olfactory perception in C1 were higher than those of C4, and the scores of each aspect of the tactile dimension in C4 were significantly higher than those of C1, so C1 was named the “Olfactory Prominent Group” (24.7%) and C4 was named the “Tactile Prominent Group” (22.2%). C2 had the highest scores in all dimensions and had a good comprehensive experience of spatial perception, so it was named the “Comprehensive Perception Excellence Group” (22.1%). Category C3 scored lower than the other three groups in all dimensions and was named the “Comprehensive Perception Deficiency Group” (31.0%).

Fig. 5: Response probability maps for each potential category.
figure 5

a Spatial perception; (b) behavioral preferences.

Full size image

Similarly, as shown in Fig. 5b, there is little difference between the four groups in terms of frequency of use and transportation modes, but a significant difference is observed in their dwell time. The dwell time for category C1 is all “over 4 h” and the number of people who use the park “daily” is much higher than other groups, so it is called the “Frequent-Visitor and Extra-Long-Stay Group” (8.0%); category C2 individuals stay “within 30 min” and the majority use the park “occasionally (or rarely)”, thus it is called the “Occasional Short-Stay Group” (16.9%); likewise, based on the results shown, category C3 is named the “Frequent Long-Stay Group” (23.2%), and C4 the “Frequent Short-Stay Group” (51.8%).

By analyzing the results of health effects, it can be seen that in general, urban parks have an improving effect on the restorative effects on the health of the population, in which 67.8% and 71. 4% of the respondents in the C1 category believe that the psychological and physiological discomfort has been slightly reduced, and therefore the C1 group is called the “Weakly Restorative Group”. And in the C2 category, 89.0% and 81.9% of the respondents believe that the psychological and physiological discomfort has been significantly reduced. Therefore the C2 group is called the “Obviously Restorative Group”.

The study further analyzed the population attribute characteristics of each spatial perception and behavioral preference group. The findings indicated that male visitors and older adults tended to visit parks more frequently and stay longer compared to female visitors and teenagers under 18. Furthermore, clerical or office workers, professionals, and retired residents were more likely to visit parks regularly. However, visitors with elementary school education visited parks less frequently and for shorter durations. Financial workers and clerical office workers exhibited stronger olfactory and tactile perceptions of parks respectively, while individuals with lower education levels may have better tactile experiences. Moreover, individuals with lower physical and mental health levels had a better overall perception but reduced the frequency and duration of park visits. Interestingly, better psychological state was associated with more prominent olfactory perception.

Mechanisms of health effects in urban park spaces

Correlation analysis

The correlation coefficient between spatial perception and health effect was 0.564, indicating a high positive correlation. This means that the higher the degree of spatial perception of the crowd in the urban park, the more noticeable the improvement in their physical and mental health. Similarly, the correlation coefficient between spatial perception and behavioral preference was 0.531, also showing a positive correlation. This suggests that the degree of spatial perception of the crowd positively influenced their choice of behavior. Additionally, the correlation coefficient between behavioral preference and health effect was 0.470, indicating a positive relationship between the two.

Reliability analysis

The reliability and validity analysis resulted in α = 0.924, and KMO = 0.884, which was >0.5 and close to 1. The P-value was 0, and the significance was less than the criterion of 0.05, indicating high reliability of the questionnaire and reasonable model structure for factor analysis. The rotated component matrix obtained through exploratory factors (Table 5) showed that the factor loadings were all >0.5, indicating significant correlation between variables in each component and good scale structure validity. In the validation factors (Table 6), the factor loadings were all >0.6 and not greater than 0.95. Additionally, the combined reliability CRs of the latent variables were all >0.7, meeting the criterion of 0.6, and the mean-variance extracted AVEs were all >0.5. Therefore, the sample data for each indicator exhibited high reliability.

Table 5 Component matrix after rotation.
Full size table
Table 6 Results of validation factor analysis.
Full size table

Hypothesis testing and analysis of model results

The parameter estimation was carried out using the maximum likelihood method for the specified model to test the validity of the relevant hypotheses. Eight selected fitting indices were within reasonable limits. However, the CMIN/DF criterion was not met. Therefore, this study explored the correlation between the residual term e2 of the tactile factor and the residual term e4 of the auditory factor to make necessary modifications and adjustments to the model. This approach can improve the stability of the residual distribution, thereby enhancing the explanatory power of the model. The final model’s all paths had a P-value <0.05 and a t-value >1.96, indicating their significance and supporting the validity of the hypotheses.The results of the analysis (Fig. 6) showed that all hypotheses were tested. We then proceeded to discuss the detailed results of each hypothesis test.

Fig. 6
figure 6

Standardized parameter estimation paths for the final model of structural.Blue arrows represent relationship paths, the width of each arrow is proportional to the strength of the relationship.*** indicate statistical significance at p < 0.001.

Full size image

Regarding the H1 hypothesis, the results indicated that the perception of the urban park spatial environment had a significant positive effect on the health recovery effect of the population, with a standardized path coefficient of 0.46. Among the four measurement items of spatial perception, the olfactory factor and the visual factor were found to have the greatest influence. This suggests that olfactory and visual factors play a crucial role in the evaluation of spatial perception. In the previous statistics, the scent of green grass and light flowers was considered comfortable by more respondents, which can effectively relieve anxiety and improve mood.

The results of hypothesis H2 indicated that people’s perception of the urban park spatial environment had a significant positive impact on their behavioral preferences, with a standardized path coefficient of 0.57. This finding suggests that individuals’ perception of the park’s spatial characteristics evokes emotions, which in turn influence their behavioral choices. Therefore, it can be inferred that people’s behavioral preferences are influenced by the resonance of their spatial perception of the park.

Hypothesis H3 was also supported, as population behavioral preferences had a significant positive effect on the population health recovery effect, with a standardized path coefficient of 0.46. Among the dimensions of behavioral preferences, the frequency of park use and the duration of park visits had the greatest impact on the choice of behavioral activities, with coefficients of 0.91 and 0.90, respectively. This implies that individuals who use the parks more frequently and spend longer periods there have greater exposure to the natural environment, leading to greater restoration benefits and positive effects on heart rate recovery.

Finally, hypothesis H4 also supports the idea that population behavioral preferences play a significant mediating role in urban park spatial perception when it comes to population health recovery. The direct impact effect, indirect impact effect, and total effect were found to be 0.46, 0.25, and 0.60, respectively (Table 7). These results suggest that behavioral preferences have a greater influence on the population’s restorative perceptions, both directly and indirectly. This could be attributed to the fact that spatial perception primarily measures people’s attitudes towards specific landscape elements, which represent their functional dependence on the spatial environment. On the other hand, behavioral preference reflects people’s emotional expression of the spatial environment, essentially expressing the way people identify with and utilize the spatial landscape element. Therefore, simply meeting functional needs may only fulfill people’s primary needs, while a deeper interactive experience is required to enhance the restoration benefits to the spatial environment.

Table 7 Standardized direct, indirect, and total effect tests between variables.
Full size table

In summary, the transmission lines of urban park recovery effects can be categorized into two paths: the direct influence path of landscape spatial perception on health recovery, and the indirect influence path of landscape spatial perception on behavioral preference and subsequently on health recovery. These findings deepen our understanding of the relationship between urban park spatial environment, behavioral preference, and population health recovery, and provide a theoretical basis for improving the design and utilization efficiency of urban parks.

Identification and analysis of landscape elements in urban parks

During the data preprocessing stage, we carefully screened and cleaned the collected landscape photos. After filtering out duplicate or unrecognizable photos, we ended up with a dataset that contained 1031 landscape photos. To further investigate the relationship between landscape elements and spatial perception and behavioral activities, we extracted the landscape elements from the dataset and conducted a correspondence analysis. The association pattern diagram shown in Fig. 7 illustrates the findings. Figure 7A reveals that spatial perception categories were not completely independent of the park landscape element variables; there was a correlation between them (χ2 = 5.17, p < 0.05). Among the landscape elements, landslides and step, shrub, and bridges exhibited the strongest correlation with auditory perception, followed by trees jungles and bonsai. In terms of olfactory perception, flowers and lawns had the greatest influence, followed by landscape pavilions. Visual perception showed a strong correlation with the categories of lakes and wetlands and plazas, while the other landscape elements had weaker correlations. When it comes to tactile perception, landslides and step showed the strongest correlation, whereas bonsai, trees jungles, and bridges had relatively weak correlations.

Fig. 7: Correspondence analysis diagram.
figure 7

a Spatial perception vs. landscape type; (b) behavioral activities vs. landscape type.

Full size image

The results from Fig. 7B indicated that there was a correlation between behavioral preferences and park landscape element variables (χ2 = 91.46, p < 0.05). Lakes and wetlands, landscape lights, bridges, trees and jungles were strongly associated with physical activity. Landslides and steps, trunk road, and amusement facilities also showed a positive correlation with physical activity. Shrubs and plazas were found to be more closely related to natural interactive activities, and they had a strong correlation with each other. However, all elements were distant from cultural interaction activities, although synthetic rockery and bonsai were relatively more relevant to cultural interaction activities. Landscape pavilions were found to have the strongest correlation with social interaction activities.

In summary, it was found that some landscapes evoke multiple senses or behaviors. For example, lakes and wetland can stimulate both visual perception and physical activity; Bonsai evoke auditory, tactile and cultural interactions; Shrubs, trees and jungles stimulate auditory perception and physical activity.

In conclusion, the study suggests that each type of landscape element plays a different role in urban parks. To enhance the experience and health benefits of park visitors, it is important to carefully consider the use of different types of landscape elements based on the specific spatial nodes.

Because the People’s Park is more comprehensive and can represent the current situation of parks in Chengdu, it is used as the specific practice base of PPGIS for this study. To further analyze the spatial environment of People’s Park, we collected a total of 1391 sensory cognitive elemental points and 1456 behavioral activity elemental points. We then conducted kernel density mapping to identify the strengths and weaknesses of the park’s space. Figure 8 shows that the Xinhai Railway-Protection Movement Monument, the West Gate Promenade, the Xinhai Railway-Protection Movement Square and the Wind Shelter were the most important elements of the park’s spatial environment, all of which were hotspots. These areas had the highest spatial quality and offered a wide range of perceptual and behavioral activities. On the other hand, the spatial perceptions around the Artificial Lake and Chrysanthemum Garden on the east side of the park were the weakest, with limited activities and mostly cold spots. Additionally, compared to other areas, the People’s Park had only one tactile hotspot area in the West Gate Promenade, with fewer distinctive tactile experiences, resulting in a lack of a core public node. Furthermore, the space for cultural and recreational activities was concentrated in the central part of the park and the west side of the West Gate Promenade, covering a relatively small area.

Fig. 8: Spatial distribution of sensory cognitive and behavioral activities.
figure 8

a Olfactory perception; (b) auditory perception; (c) visual perception; (d) tactile perception; (e) social interaction activities; (f) natural interactions; g cultural recreation; (h) sports.

Full size image

To address these issues, we recommend that the People’s Park integrate the different functions of various landscape elements to redesign the lakeside area and the Chrysanthemum Garden space. Additionally, the park should consider adding tactile core nodes and areas dedicated to cultural and recreational activities. These enhancements would provide visitors with a wider range of behavioral activity choices and enhance the overall appeal of the park.

Discussion

Socioeconomic attributes and perceived behavioral differences

The results of this study show that there is significant variability in perceptions and behaviors across socioeconomic groups. Specifically, the “Overall Perceived Excellence Group” was dominated by the low health group. They rated their perceptions of the park environment higher than the other groups, but their frequency and duration of visits were relatively lower. This may be attributed to the proposition that the low health group is constrained in their park visiting behavior by lower levels of health or limited ability to travel. However, such populations are more oriented to the healing effects of natural spaces, and thus their limited experience of park environments produces higher subjective perceptual ratings(Wang et al., 2025). The study also found that those with lower levels of education had a more positive experience of touch. This may be due to the fact that, as a primary and more direct perception, the less educated group relied more on tactile basic senses to interact with the environment, such as direct contact with water and plants through the skin, and that these stimuli do not require deeper cognitive interpretations(Kim et al., 2023). Not only that, the female and adolescent groups visited significantly less frequently than the male and middle-aged and elderly groups, which is consistent with previous findings. Influenced by the social and cultural background of China, females take on more responsibility for family care, while adolescents are influenced by their academic tasks, thus limiting their frequency of park visits(Zheng, 2021). In addition, women may be more concerned about safety factors, with factors such as lighting conditions and the risk of crime occurring affecting women’s visits outside the home(Li et al., 2022).

Compound pathways of health effects in urban parks

A pivotal discovery of this research is the identification of compound pathways through which urban parks impact health, specifically the “spatial perception-health recovery” and “spatial perception-behavioral preference-health recovery” pathways. This insight signifies a substantial advancement in studies examining the health effects of urban parks, diverging from prior research that primarily focused on the direct relationship between park environments and health outcomes (Holy-Hasted et al., 2022).

The research results indicate that spatial perception in urban parks has a direct and significant positive impact on health recovery effects. The study explores the interaction among various sensory factors, identifying olfactory and visual factors as the most influential, followed by auditory and tactile factors. The pronounced impact of olfactory factors is corroborated by the “Olfactory Prominent Group” within the spatial perception classification. According to studies published in the esteemed academic journal Neuron, olfactory perception is the most enduring to forget in sensory memory. The two most scents identified in this study’s olfactory factors are grass and light floral aromas, as these pleasant smells are known to alleviate anxiety and enhance mood (Xia, 2018). For instance, after exposure to the scent of gardenia flowers, diastolic blood pressure decreased from 79.783 ± 0.390mmHg to 77.933 ± 0.365 mmHg, and heart rate decreased from 83.467 ± 0.783 b/min to 81.950 ± 0.786b/min (Cui et al., 2023). This evidence supports hypotheses H1, H2, and H3 of this study.

Consistent with H4, individuals’ spatial perception experiences can influence their behavioral preferences, which in turn indirectly affect health recovery outcomes; however, the mediating effect is weaker than the direct effect. The factors impacting the extent of health effects, in descending order, are: frequency of use, dwell time, transportation mode, and activity type. This aligns with earlier classification results, which grouped participants into categories such as the “Frequent-Visitor and Extra-Long-Stay Group” and the “Occasional Short-Stay Group” nothing that there were no significant differences in activity types among these groups. A more favorable overall sensory experience, correlates with a higher frequency of park use, longer dwell times, and greater exposure to the natural environment, all contributing to improved health recovery effects, including enhancements in heart rate, blood pressure, and heart rate variability (Liu, 2022; Wang et al., 2022). These findings suggest that the frequency of use and dwell time in park green spaces significantly impact health recovery effects (P < 0.05), consistent with results from related studies(Matsuba et al., 2011; Ren, 2022). Furthermore, it indicates that the health recovery benefits for the “Frequent Long Stay Group” may be more pronounced. The mode of transportation reflects the accessibility of park green spaces. If people can spend less time and material resources to reach the park, the relative accessibility increases, leading to better health recovery effects (Fernandez et al., 2024).

Our comprehensive examination of these compound pathways has successfully substantiated hypotheses H1, H2, and H3, establishing a significant correlation between the environmental perception of urban parks and their health restoration effects. More critically, our research has illuminated the pivotal mediating role of behavioral preferences in this relationship (H4). This nuanced understanding enhances the knowledge of how urban park design influences public health and establishes a novel theoretical foundation for future research in this domain.

From a practical standpoint, this understanding of the composite pathways offers valuable guidance for urban park planning and design, emphasizing the importance of fostering positive visitor behaviors to optimize health restoration benefits (Motomura et al., 2022). This approach not only enriches theoretical discussions but also serves as an innovative and scientific reference for urban park planning and management globally, facilitating the creation of health-promoting public spaces.

Multi-sensory natural landscape elements: enhancing health outcomes in urban parks

Based on the composite path of health recovery effects, it is essential to further implement the transformation of landscape elements within the park’s spatial environment. The multisensory cognitive results indicate that natural landscape elements in urban parks are more likely to capture attention without inducing stress, thereby mitigating negative effects (Ruotolo et al., 2024). These natural elements can demonstrate restorative effects by fostering positive emotions, which facilitate recovery from stress(Ulrich et al., 1991). This finding aligns with “Attention Restoration Theory” (Kaplan et al., 1989) and “Stress Recovery Theory” (Joye et al., 2011).

This study illuminates the crucial role of natural landscape elements—such as trees, jungles, shrubs, lakes, and flowers—in urban parks, particularly through multisensory pathways that promote visitors’ health recovery. Our findings provide a more comprehensive perspective than previous research, emphasizing the importance of these natural features in enhancing auditory, visual, tactile, and olfactory experiences, which contribute to both physical and mental health benefits (Alvarsson et al., 2010; Ba et al., 2022; Ruotolo et al., 2024).

For example, trees and jungles create tranquil and secluded environments that are conducive to physical activity and health restoration (Buchel et al., 2015). However, our research also indicates that excessively dense vegetation may raise safety concerns, underscoring a critical consideration for urban park planning (Falk et al., 2010). Therefore, the optimal density and distribution of vegetation are essential for achieving a balance between aesthetic appeal and functional utility.

Lawns serve as critical visual elements that significantly enhance visitors’ sense of comfort and belonging (Jin, 2022). While they help alleviate psychological stress and promote social interactions, overly dense vegetation can obstruct spatial visibility, potentially diminishing these benefits (Roovers et al., 2006). Thus, thoughtful plant arrangement and spatial openness are essential considerations in park design.

Water features in lakes and wetlands are particularly effective for mental health recovery, providing serene and comforting environments (White et al., 2010). Our study indicates that the diversity and interactivity of water-based experiences are fundamental to maximizing this restorative impact (Ratcliffe et al., 2016). Designs that limit tactile interactions with water may inadvertently diminish the potential health benefits.

Moreover, the study highlights that the collective value of these natural landscapes lies in their capacity to offer aesthetic, natural, and spiritual experiences (Shoyama et al., 2016). ements such as gentle slopes, steps, and bridges (Hong et al., 2013; Nielbo et al., 2013), along with the aromatic allure of flowers (Cui et al., 2023), combine to create a rich multi-sensory experience that attracts visitors and enhances wellness through diverse sensory engagements.

Addressing research limitations and future directions

This study, while providing valuable insights, acknowledges several limitations:

  1. I.

    Reliance on Self-Reported Health Assessments: The health effect assessments in this study were based on self-reported data from respondents, which could introduce biases or inaccuracies compared to objective measurements like electroencephalography or electrodermal activity testing (Li et al., 2023). Future studies might benefit from incorporating these objective testing methods to validate and complement self-reported findings.

  2. II.

    Focus on People’s Park Spatial Optimization: The current research centers on the spatial optimization of People’s Park, a comprehensive park. Although phased results have been achieved, it is unfortunate that the spatial status and challenges of other types of parks have not been extensively explored. This limitation may hinder the ability to provide a comprehensive and representative picture of the issues faced by various types of parks. Therefore, future work needs to be further refined, adopting a strategy of addressing each park type individually, and conducting in-depth research and implementation of the renovation needs specific to different parks to ensure the comprehensiveness and relevance of park space optimization.

  3. III.

    Differential Impacts on Various Population Groups: While the study examined the overall effects of landscape elements on spatial perception and behavioral activities, it did not investigate the specific impacts on different demographic groups. Future research could explore how diverse groups experience and benefit from urban parks differently, potentially uncovering nuanced understandings of health recovery processes across various populations.

By addressing these limitations in future studies, the field can advance towards a more comprehensive and nuanced understanding of the health effects of urban parks. Integrating objective health measures, leveraging technology for data analysis, and focusing on demographic-specific impacts will enhance the depth and applicability of research findings in this area.

Conclusions and implications

This research sought to address the limitations inherent in traditional questionnaire methods, which are often inadequate for effectively capturing dynamic multisensory cognitive and behavioral data. By integrating the Public Participation Geographic Information System (PPGIS) approach, we adopted a novel perspective to comprehensively analyze variances in multisensory cognition, behavioral preferences, and the health effects of urban parks across diverse populations. This study aimed to elucidate the complex mechanisms through which urban parks’ spatial environments contribute to public health recovery and to identify key landscape elements that facilitate this process.

Our findings revealed marked differences in spatial perception, behavioral preferences, and health restoration effects, largely attributed to the socio-demographic attributes of the population. For instance, male and older adults were more inclined to frequent parks and for extended durations. While professionals and retirees showed a higher tendency to visit parks, those with lower education levels demonstrated lesser propensity. Additionally, individuals with lower physical and mental health were found to visit parks less often and for shorter durations, yet those with better mental health reported heightened olfactory perception.

Furthermore, the study uncovered that the health impact mechanisms in urban parks operate via two primary pathways: a direct influence pathway (landscape spatial perception-health recovery) and a mediated pathway (landscape spatial perception-behavioral preference-health recovery). In the restoration process, auditory perception was most strongly linked to elements like gentle slopes, steps, shrubs, and bridges. Olfactory perception correlated prominently with flowers, lawns, and landscape pavilions. Visual perception showed a significant association with lakes, wetlands, and plazas, while tactile perception aligned closely with gentle slopes and steps. Regarding behavioral preferences, physical activity was most associated with lakes, wetlands, landscape lights, and nature interaction correlated closely with shrubs and trail plazas. Cultural interaction was predominantly linked with man-made features like wigwams and potted plants, and social interaction with landscape pavilions.

To effectively cater to the diverse needs of the public, we recommend enhancing urban park designs to include a broader range of multisensory stimulation activities and facilities, encompassing visual, auditory, olfactory, tactile, and gustatory experiences. Additionally, it is vital to design spatial environments that align with the behavioral preferences of various demographic groups, integrate a diverse array of landscape elements, and leverage innovative technologies like PPGIS to augment park management and operational efficiency.

These insights hold substantial implications for optimizing the spatial environment of urban parks and for the development of park city demonstration areas in Chengdu. They not only contribute to the enrichment of urban park design and public health research but also offer novel approaches and methodologies for the optimal design of urban parks.