Health perception and restorative experience in the therapeutic landscape of urban wetland parks during the COVID-19 pandemic

Introduction The effects of restoration and inspiration in the therapeutic landscape of natural environments on visitors during the COVID-19 pandemic have been well-documented. However, less attention has been paid to the heterogeneity of visitor perceptions of health and the potential impacts of experiences in wetland parks with green and blue spaces on visitors’ overall perceived health. In this study, we investigate the impact of the restorative landscapes of wetland parks on visitors’ health perceptions in the context of the COVID-19 pandemic. Methods In our survey, 582 respondents participated in an online questionnaire. We analyzed the respondents’ health perceptions in terms of latent class analysis, used multinomial logistic regression to determine the factors influencing the potential categorization of health perceptions, and used structural equation modeling to validate the relationships between health perceptions of different groups and landscape perceptions of wetland parks, restorative experiences, and personality optimistic tendencies. Results The results identified three latent classes of health perceptions. Gender, marital status, education, occupation, income, distance, frequency of activities, and intensity of activities were significant predictors of potential classes of perceived health impacts among wetland park visitors. Discussion This study revealed the nature and strength of the relationships between health perception and landscape perception, restorative experience, and dispositional optimism tendencies in wetland parks. These findings can be targeted not only to improve visitor health recovery but also to provide effective references and recommendations for wetland park design, planning, and management practices during and after an epidemic.


Introduction
The COVID-19 pandemic has had a profound impact on human society (1).Although control measures have calmed the outbreak, they have also exacerbated mental health problems (2, 3).A growing body of literature underlines the importance of extended psychological and physical health distress during the COVID-19 epidemic (4,5), which is considered a public health concern.Although people may suffer from many adverse effects on their health during an epidemic, such as fear of viral infection, prolonged isolation, and other complex factors, people with optimistic dispositions tend to ease stress and anxiety caused by the epidemic because of their positive psychological cues, improve their physical health, seek social interactions to fulfill their social roles, and maintain their intrinsic mental and emotional health (6, 7).
Natural areas can provide an escape from mundane urban life (8).Many studies have suggested that natural connectivity in a therapeutic landscape provides a comfortable and healing environment for tourism and recreation (8)(9)(10).This reduction in mental and physical resources can be offset by access to a therapeutic environment (11,12).Thus, visits to wetland parks, which are special parks with therapeutic landscapes in urban areas, may effectively prevent anxiety and depression caused by short-and long-term quarantines and lockdowns in natural environments (13)(14)(15).Wetland parks in urban areas improve water and air quality, support biodiversity, and promote human comfort (16,17).However, less attention has been paid to the important role of therapeutic landscapes in urban wetland parks in healing, restoring, and maintaining human health (18)(19)(20).Therefore, it is paramount to understand how the experience of wetland parks benefits visitors' perceived health.This study attempts to broaden the knowledge of how the experience of wetland parks' therapeutic environment affects people's health perceptions.
In addition, evidence shows differences in individuals' perceptions of their health status based on gender, age, personality traits, and cultural background.These heterogeneities can directly influence visitors' attitudes and behaviors toward health issues.Highly preferable and stimulating landscapes in wetland parks can improve people's mental state and promote physical wellbeing.However, factors such as social background and personal preferences differ between individuals, leading to differences in their subjective perceptions of health.Therefore, understanding the heterogeneity of visitors' health perceptions during their visit to wetland parks is important for improving their health status.Heterogeneity in health perceptions has become an important issue in several fields, including social psychology and behavioral medicine.However, academic research on health perception in wetland parks is lacking.Therefore, further research is needed to understand the heterogeneity in the health perceptions of wetland park visitors and guide the planning and management of wetland park landscapes.
The following section proposes a comprehensive framework through which we explore the nature and strength of the relationships between people's perceived health and their experiences with the therapeutic landscape in terms of landscape perception, restorative experiences, and dispositional optimism tendencies in wetland parks.This is followed by a description of the methods used for data collection and modeling approaches, including latent class analysis (LCA) and structural equation modeling (SEM).Based on these results, the key findings and discussion are presented in Section 4. Finally, conclusions and implications are presented in the last section.

Conceptual framework and hypothesis
The landscape comprises physical, social, and symbolic aspects (21).People perceive the physical landscape as the interaction between individuals and the natural and built environments (22).The park, which is characterized by highly visible mowed areas, open water, and flourishing planting mixtures, provides a diverse and picturesque landscape for visitors to enjoy (23)(24)(25)(26)(27).It is attractive to walkers, dog walkers, and cyclists and provides space and opportunities for physical activity (28,29).The social aspect of the landscape is perceived by individuals when they interact socially, and it includes a wide range of themes related to public attitudes, values, behaviors, and activities (30).The social aspect of the landscape benefits social integration by providing public spaces for activities that generate broad social connections and overcome people's feelings of social isolation and exclusion (31,32).The symbolic landscape, as a psychological cue, is associated with the public's belief in the healing characteristics of the therapeutic environment (33)(34)(35)(36).For example, national parks in the United States function prominently in American culture and religion, with symbolic meanings of 'inspiration' , 'stability' , and 'healing' (37).
By visiting urban wetland parks, people can access the therapeutic landscape of nature with water, diverse vegetation, open lawns, and bright flowers, resulting in substantial improvements in mood, cognitive function, and mental health (38)(39)(40).A temporary escape from the dreary pressures of work and life and immersion in nature can satisfy the need for fresh air, sunlight, water, and food, thus facilitating a restorative experience (41).Moreover, comfortable public spaces in wetland parks allow visitors to interact socially, independent of their socio-economic status (31,32).Wetland parks provide uncrowded and undisturbed spaces for visitors seeking solitary experiences, such as meditation and independent thinking, in natural environments.In addition, visitors can achieve psychological benefits through horticultural activities that ultimately improve their perception of restoration (40).
According to the literature, an individual's health is influenced by environmental factors, sociodemographic characteristics, and psychological traits (18,39,42,43).Dispositional optimism is a positive psychological trait often associated with good health and relief from stress and anxiety (44).Therefore, experiences in wetland parks may have an impact on health perception, which may vary among individuals with various sociodemographic backgrounds and behavioral activities.
Previous studies have indicated a diversity in individuals' health perceptions (45,46).However, it cannot be simply determined a priori from the relevant observed variables because the unobserved heterogeneity in an individual's perception is not necessarily captured by the variables preconceived and specified by existing theoretical and conceptual models, but it can exist outside of the previously identified variables (47).For example, the results of various studies on gender differences in health perceptions in wetland parks are mixed: Some studies have found no significant differences in health perceptions between men and women who visited wetland parks before and after the outbreak (39), while others have shown that women who visited nature reserves felt healthier than men (48).It is imperative to classify heterogeneous visitors into multiple homogeneous groups and identify appropriate treatments for wetland park planning and design.A new modeling framework (Figure 1) was developed to accurately capture the health behavior patterns and core characteristics of different populations.

Methods
LCA is a finite mixture model.It is a statistical method that allows associations between exogenous indicators to be explained by an intermittent latent class so that associations between exogenous indicators are explained by the latent class and thus maintain their local independence (49, 50).LCA is used to identify unobservable classes within a population that may have similar health perceptions.It provides a medium through which results can be expressed in terms of probabilities rather than fixed deterministic conclusions (49, 51).The relationships between individuals' landscape perceptions, restorative experiences, and dispositional optimism tendencies were analyzed separately for different classes of visitors.
The mathematical expression for LCA is shown in the following equation: where The latent class probabilities are to satisfy equation (2).
Conditional probability reflects the strength of the relationship between latent and exogenous variables.The formula used is as follows: SEM is a general statistical approach used to assess the relationships between observed and latent variables (52), and it can estimate specific path coefficients based on identified latent classes.
The equation of the measured model is as follows: where equations ( 4) and ( 5) are the exogenous and endogenous indicators, respectively.Λ is the relationship between inventory and latent variables, and δ and ε are measurement errors.
The structural model is a path diagram reflecting the relationship between the effects of potential variables and is formulated as follows: where η is the endogenous latent variable, ξ is the exogenous latent variable, and Β denotes the effect of the exogenous latent variable on the endogenous latent variable.Γ denotes the effect of some endogenous latent variables on other endogenous latent variables, and ζ is the regression residual.

Data collection
The questionnaire consisted of seven sections: (I) sociodemographics, (II) physical landscape perception, (III) social landscape perception, (IV) symbolic landscape perception, (V) restorative landscape perception, (VI) health perception, and (VII) the C-WEMWBS Warwick-Edinburgh Positive Psychology Scale.The Health Perception Scale was derived from Ware's General Health Perception Scale (53).The Restorative Experiences Scale was derived from Korpela (54), Korpela (55), and Von Lindern et al. (56).The Dispositional Optimism Scale was derived from the Warwick-Edinburgh Mental Health Scale (WEMWBS).Questionnaires were distributed across China via an online platform on 15 February 2022.Before the official questionnaire was distributed, a sample of wetland park visitors was taken for a pre-survey to verify its suitability and check for any ambiguities so that the questionnaire could be revised and improved.The survey was continued only if the respondents indicated their experience of visiting a wetland park.Table 1 shows the questionnaire form.

Descriptive statistics
Table 2 presents the socio-demographic characteristics, socioeconomic status, and dwelling conditions of the respondents.Slightly more female visitors than male visitors participated in the survey.Of the respondents, 18-45-year-olds were the majority, followed by those aged 45 years.In terms of marriage, 63.1% of the visitors were unmarried, and 36.9% were married.Approximately 76.1% of the respondents had a bachelor's degree.Students were predominant in the sample population, accounting for 34.9% of respondents, while employed individuals accounted for 34.2%.The respondents' monthly income is denoted in Chinese Yuan (CNY), and the income of more than one-third of the respondents did not exceed 3,000 CNY.Regarding housing tenure status, most respondents lived in dormitories or with their families, and approximately 70% of respondents did not own property.Most of the respondents had lived in the local area for more than 3 years.Nearly half of the respondents reported that their neighborhoods were under semi-containment or full containment as a result of the outbreak.Of these, 13.1% were strictly confined to their homes during the outbreak and could not move freely within their neighborhoods.In addition, 44% of them were required to record their temperature when entering and leaving the neighborhood.
Table 3 lists the factors related to transportation.More than half of the visitors traveled more than 5 km to the wetland park, and only 8.4% traveled less than 1 km; thus, the wetland park attracts many visitors who live far from it.In terms of transport choice, 41.6% of visitors chose to travel by taxi or car, with the fewest visitors choosing to travel by motorbike or bicycle, accounting for 11% of the sample size.The respondents who visited wetland parks once every 6 months accounted for 47.1% of the sample size, and only 2.6% visited wetland parks weekly.For most of them, the main purpose of visiting was to rest and relax.In terms of activity intensity, 84.2% of visitors chose low-intensity activities, whereas only 2.1% engaged in high-intensity activities.Regarding the form of accompaniment on a visit to a wetland park, 45.5% of visitors indicated that they usually visited a wetland park with friends, followed by family members, accounting for 39.7% of the sample size (Table 4).

Identifying distinct classes of health perception
Models with increments in the number of classes were estimated to detect a suitable number of latent classes.Multiple statistical indicators were used to determine the best-fit model.The smaller the values of the Akaike information criterion (AIC), Bayesian information criterion (BIC), and adjusted Bayesian information criterion (aBIC), the better the model's fit (47,57).Entropy evaluates the accuracy of the model classification, and a value closer to 1 indicates a better fit (58).The Lo-Mendell-Rubin

PLP01
The wetland park has a large area of natural and human-made lakes and reservoirs

PLP02
The wetland park is rich in animal and plant resources

PLP03
The trees in the wetland park provide shade for pedestrians

PLP04
The wetland park has enough open space and water-friendly leisure facilities

PLP05
The wetland park has walkways for people

PLP06
The wetland park provides a place to breathe fresh air

PLP07
The wetland park provides a place to soak up the sun

PLP08
The wetland park provides a place to observe nature and appreciate the beauty (LMR) and bootstrapped likelihood ratio test (BLRT) provide a value of p that can be used to compare the increase in model fit between neighboring class models and to determine a statistically significant improvement in fit when another class is included.As shown in Table 5, the AIC, BIC, and aBIC decreased as the number of categories increased.It is possible that the higher the number of categories, the better the fit.However, from Class 3, the entropy started to decrease, and from Class 4, the LMR (P) also started to become statistically insignificant (p > 0.05), indicating that the model's fit gradually deteriorated from Class 3. The entropy was closest to 1 in Class 3, which means that the model has the best-fit and classification accuracy.A model with three latent classes was selected as the most appropriate for this study, considering the practical significance of the classes and the fit values p represented by the classes.
Based on the three-class model, Figure 2 illustrates the probabilities of respondents' health perception in each latent class.Latent class 1 (C1) was characterized by high probabilities for all health perception components and labeled 'good self-perception of health' (n = 323, 55.5%).Latent class 2 (C2) was characterized by high probabilities of good health status in the present and future but did not improve significantly after the visit to the wetland park and was, therefore, named 'neutral self-perception of health' (n = 192, 33.0%).Latent class 3 (C3) was characterized by low probabilities of all health perceptions; accordingly, we labeled it 'poor self-perception of health' (n = 67, 11.5%).

Roles of different factors in predicting class membership
We further conducted a multinomial logistic regression model to examine the respondents' health perceptions associated with their social backgrounds and behavioral activities, with Class 2 (medium self-perception of health) as the reference group (Table 6).
Compared with Class 2, respondents in Class 1 (good selfperception of health) were more likely to be unmarried (OR = 1.24, p < 0.001), master and above (OR = 1.25, p < 0.001), students or freelancers (OR = 1.22,p < 0.001; OR = 1.36, p < 0.01), in a group accommodation (OR = 1.22,p < 0.05) and under open and semi-closure management of residential areas during the outbreak period (OR = 1.21, p < 0.05; OR = 1.21, p < 0.05).The main purposes for visiting the park were more likely to be for rest and relaxation (OR = 1.25, p < 0.01).The probabilities of visiting the park with friends were high, and weekly park visits before the outbreak were high (OR = 1.21, p < 0.05; OR = 1.86, p < 0.001).
Compared with Class 2, respondents in Class 3 (poor selfperception of health) were more likely to be female (OR = 1, p < 0.05), unmarried (OR = 1, p < 0.001), with a higher vocational or junior college degree (OR = 1.07, p < 0.05), with a master's degree and above (OR = 2, p < 0.05), students (OR = 1, p < 0.05), in a group accommodation (OR = 1, p < 0.05), and under semiclosure management of residential areas during the outbreak period (OR = 1, p < 0.01).Accommodation was more than 5 km away from the wetland park (OR = 1.84, p < 0.05).The main purposes of visiting the park were more likely to be for rest and relaxation (OR = 1, p < 0.001).The probabilities of visiting the park with friends and visiting the park 6 months before the outbreak (OR = 1.36, p < 0.05; OR = 1, p < 0.05).

Structural equation modeling results
Table 7 shows that the critical fitting indices of C1 and C2 are within the acceptable recommended value range; thus, models C1 and C2 have good fitting degrees.The CFI and TLI of C3 are within the acceptable recommended value range, respectively; however, the RMSEA of C3 suggests an insufficient model fit.Overall, the model's fit was considered acceptable, as RMSEA can suggest a below-adequate model fit for models with low degrees  The research model's hypothesis was validated using AMOS 24.0, and the standardized path coefficients and significance levels between the latent variables are shown in Figure 3.The final hypothesis validation results are shown in Figure 3.
As shown in Tables 8-10, the standardized factor loadings between the latent variables and all the corresponding observed variables are >0.5.Regarding the structural models, Table 11 presents the estimation results for the three latent classes; those for latent class 1 are as follows (Figure 3A).Visitor's physical landscape perception (γ = 0.114, p = 0.021) and social landscape perception (γ = 0.285, p <0.001) were positively significant on health perception.Visitors' perceptions of the social landscape (γ = 0.191, p = 0.005) and symbolic landscape (γ = 0.350, p < 0.001) were positively significant for their restorative experiences.The effect of visitors' restorative experience (γ = 0.332, p<0.001) and character optimism (γ = 0.378, p<0.001) on health perceptions was positively significant.The relationship between the visitors' symbolic landscape perception variables (γ = 0.039, p = 0.526) and health perceptions was not significant.

Discussion
We investigated the impact of restorative landscapes in wetland parks on the health perceptions of respondents during the COVID-19 pandemic.By classifying respondents' health perceptions into latent classes, we found that visitors could be classified into three groups with significant group heterogeneity in the population.Sociodemographic characteristics and behavioral factors showed different associations with each group.
Compared with the 'medium self-perception of health' class, respondents in the 'good self-perception of health' group were more likely to be freelancers and have a higher education level.Consistent with previous studies, freelancers have more time at   results also showed that residents of unsealed neighborhoods had a higher probability of belonging to the 'good self-perception of health' category during the peak of the epidemic.Lockdown measures have been a panacea for pandemic control.However, home restrictions and the overall disruption of personal daily life have made people exercise less; the workspace in a home office environment has increased the chances of physical pain and other physical health conditions, and blurred work-life boundaries can make it difficult to detach mentally from work, which can increase stress and anxiety.Therefore, when dealing with major safety and health events such as the epidemic, the community and the wetland park should adopt a 'closed ' and 'open' relationship (12, 60-64).The frequency and intensity of park cleaning and disinfection can be increased within the park, and infrastructure to stimulate exercise can be incorporated.The maximum number of visitors to a park during an outbreak can also be predicted, and spaces with a large distribution of visitors can be monitored for flow and effective evacuation guidance.Outside the parks, it is necessary to increase connections with residents in neighboring communities and improve residents' accessibility to parks (65).
Compared to the 'medium self-perception of health' group, the 'poor self-perception of health' group was more likely to include women.Consistent with previous studies, gender strongly shapes the experience of visitors to urban parks, with men being more likely than women to rate health benefits when visiting urban blue-green spaces at peak times.In addition, the more frequently residents visit parks, the better their perceived health status, which indicates that park users develop place attachment and increased affinity with nature, which contribute to health benefits.The more residents lived in a wetland park, the worse their perceived health.This is because, at the peak of the epidemic, many non-essential commercial and public spaces were off-limits.People's need to visit outdoor spaces was higher than ever.Thus, the closer the urban wetland park was to the neighbourhood, the more people had access to the park, and the more likely they felt healthy.Therefore, it is possible to extend the opening hours of wetland park services during epidemics and improve accessibility to parks while observing more accurate infection prevention methods to ensure wetland park use (65,66).
The SEM results of the three categories showed that landscape perception, optimistic personality tendencies, and restorative experiences had different strengths and associations with health perception.It is worth noting that in the first category, the symbolic landscape has no positive effect on health perception and restorative experience, possibly because the environmental elements and cultural cues associated with symbolic restorative health in China's wetland parks have not yet established positive images for people.Fewer psychological hints about health are available to visitors compared to those available at famous sites known for longevity and healing.Therefore, in future construction of wetland parks, planning designers can try to add cultural clues and symbols with symbolic significance for health and create environmental factors conducive to the healthy restoration of wetland parks.Additionally, the results showed that optimism had an impact on health perceptions regardless of group, and optimists' positive expectations are not limited to specific areas of behavior or types of circumstances.

Conclusion
This study examined the impact of wetland park landscapes on different health-perceiving populations during the COVID-19 pandemic.The study identified various factors affecting the potential categories of health perception of visitors, such as gender, marital status, education, work status, housing, community management, distance from home, purpose, form of companionship, and frequency of pre-pandemic park visits.The existence of different pathways and correspondence coefficients among the three potential categories of visitor groups was also investigated.In the event of a future major safety and health event, wetland park managers should pre-emptively address the need for 'closed' and 'open' relationships between communities and wetland parks and try to increase cultural cues and symbols with provides a relaxed social environment SL02 The wetland park provides a place for face-to-face communication SL03 In the wetland park, I can communicate and interact with my peers more actively SL04 I like to walk my dog or take my baby to the wetland park SL05 It is easier to find like-minded people with similar interests in wetland parks SL06 The wetland park promotes the development of restaurants and real estate around it Symbolic landscape perception (SLP) SLP01 The wetland park is rich in aquatic plants SLP02 Wetland landforms are rich in types SLP03 The wetland park can provide freshwater resources, aquatic products SLP04 Known as the 'green lung of the city' , the park boasts a wide variety of plants SLP05 Wetland parks reduce urban carbon emissions SLP06 Wetland parks can mitigate runoff, store flood and prevent drought SLP07 Wetland parks have open water and are known as the 'kidneys of the Earth' SLP08 Wetlands are called 'cradles of species' Restorative experience (PE) PE01 In the wetland park, it can solve my mental health problems and psychological distress PE02 In the wetland park, I can experience harmony between humans and nature PE03 In the wetland park, I can get rid of the dull work pressure and life pressure temporarily Health perception (HP) HP01 My health is now excellent HP02 I do not think I'm ill HP03 I do not think I'm in sub-optimal health at the moment HP04 I feel better when I'm in the wetland park than when I'm not HP05 In the near future, I expect to have better health than other people I know HP06 I will not probably be sick a lot in the future HP07 I expect to have a very healthy life HP08 In the future, I will probably have fewer health problems than most people around me HP09 After visiting the wetland park, my sleep quality improved HP10 After visiting the wetland park, my breathing became more relaxed HP11 After visiting the wetland park, I had more energy HP12 After visiting the wetland park, I became more active HP13 After visiting the wetland park, my reflexes became stronger HP14 After visiting the wetland park, I look better HP15 After visiting the wetland park, my physical pain disappeared or lessened

TABLE 5
Estimated indices for comparing different numbers of latent classes.
FIGURE 2Conditional probability of three-class model.their disposal and fewer stressors than other professionals, and participating in leisure and recreational activities can reduce psychological stress and improve their physical and mental wellbeing.With the improvement in education level, highly educated people have a relatively high preference for wild nature and a strong restorative experience in wetland parks (23).Our

TABLE 6
Results of multinomial logistic regression., dominance ratio; if the dominance ratio is greater than one, the probability of membership being classified into a particular class increases with each unit of the predictor variable.If the dominance ratio is less than 1, the probability of membership being classified into a particular class decreases for each unit of the predictor variable. OR

TABLE 7
Structural equation model fit index values.

TABLE 8
Results of the measurement model of C1. **p < 0.01, **0.01 ≤ p < 0.05, *0.01 ≤ p < 0.1.However, this study had several limitations.First, the findings were limited to Chinese participants.Future studies should explore cross-cultural comparisons among visitors from different countries.Second, this study did not examine the dynamics of health perceptions over time.Future research could examine longitudinal changes in the health perceptions of wetland parks.These insights can inform the decision-making process and contribute to better planning and management of wetland parks. *

TABLE 9
Results of the measurement model of C2.