Located at the Yangtze River estuary in eastern China and facing the Pacific Ocean, Shanghai is the largest Chinese economic center and also a leading global financial center. By the end of 2023, the administrative area of Shanghai was 6340.5 km², with 2.49×10⁷ permanent residents and a per capita disposable income of 84834 yuan per year. Shanghai has 16 districts, including seven central districts. The population density in the central city is high and the land supply is tight. The mean area of park available to urban residents is smaller than in the suburbs (Figure 1).

According to the Shanghai Municipal Bureau of Greening and City Appearance (https://lhsr.sh.gov.cn/), Shanghai’s seven central districts house 152 parks. Six green parks in urban residential areas of the central city were chosen as the case sites, covering areas of 2-3 ha (Yichuan Park and Jingan Park), 4-9 ha (Kangjian Park and Fuxing Park), and about 20 ha (Zhongshan Park and Luxun Park). The selected parks have various features, including children’s play areas, walkways, benches, and open green spaces for sports activities or relaxation. Certain parks have fewer functions and more limited green space, while others have more aesthetic landscapes and broad green spaces. The selected parks are situated in populous areas, have been open for at least a decade, and are available to the public for free every day. These six parks are surrounded mainly by residential buildings, but also some commercial buildings. Figure 1 illustrates the positions, distribution and characteristics of these six parks.

Children aged 6-9 were taken as research object for two reasons. First, this age group is a critical period for the formation of children’s perception and behaviors, and it is also a period where changes can be made relatively easily (McFarland et al., 2014); and second, the children’s own factors have relatively little influence on their use of green parks (Truong et al., 2023).

The data were collected by questionnaire, and the respondents were parents who take their children to play in the parks. In general, parents are the adults with the greatest influence on their children, which is why only the parents of children in the target age group were taken as respondents. In the questionnaire, parents were asked about their perceived benefits from children’s park activities, their worries about the potential for accidents, their working hours, their COE, and the frequency with which they take their children to the park, as well as their educational background, gender, and income.

Previous studies have determined that the accessibility of green spaces is an important factor affecting green space utilization. To control for the influence of green space accessibility on park space utilization, only children whose residence was within a 500-m park periphery were included in this study (Wang et al., 2021).

Based on the above criteria, three screening questions were used in this study. The first question was “How old are your children”, and only data on children aged 6-9 were used for analysis. The second question was “Are you the child’s parent(s)?”, and only data from the child’s parents were used for analysis. The third question was “How many meters is your home from the park”, and only data from children whose home was less than 500 m from the park were used for analysis. Only those responses that met all three screening conditions were used for subsequent analysis.

This study was conducted from March to June in 2023, and it was conducted at different times on both weekends and weekdays, using convenient sampling methods. The respondents were informed of the research objectives, and those agreeing to participate in the survey were provided with a questionnaire. The respondents were guided to finish the questionnaire, and 1000 questionnaires were distributed. From 160 to 170 questionnaires were distributed in each park. A total of 352 questionnaires met the three screening conditions, and 317 valid questionnaires were used for the subsequent analysis. Thirty-five questionnaires were excluded due to missing answers. Respondents did not receive any form of compensation for their responses.

Parents were asked to report the frequency of their children’s visits to that green park in the last month to measure the frequency of their children’s use of green parks. This specific question was used because asking for recent previous events can reduce memory errors and the burden on respondents (Neuvonen et al., 2007), compared with asking for the average frequency of children’s visits to green spaces in a specific period (such as one year).

In this study, the parents’ perceived benefits from their children’s green park activities were also measured. Respondents were asked to rate the statements of four benefits from green park activities on a five-point scale (from 1=totally disagree to 5=totally agree). The questionnaire was adapted from previous research (Truong et al., 2022), and the average value of benefits was used. Internal consistency was considered acceptable with Cronbach’s α=0.873.

The extent to which parents worry about the potential for accidents was assessed by asking respondents to rate the probabilities of three possible accident scenarios (i.e., attacks by strangers, accidents, and animal attacks) in green parks on a five-point scale (from 1 to 5). Internal consistency was considered acceptable with Cronbach’s α=0.855. The average values of benefits and accident worries were used to analyze the influencing factors.

To measure the parents’ COE, respondents were asked to recall how often they participated in outdoor activities and visited green spaces in their own childhood (under the age of 12 years). Outdoor activities included three main types: 1) catching fish, frogs, or insects; 2) playing with grass, flowers, and fruits; and 3) climbing trees or playing tag/hide-and-seek with a tree. These outdoor activities were very common throughout the childhood period of the parents of this generation. Respondents graded their answers according to a five-point scale (1=none at all, 2=less than once a year, 3=around once a year, 4=around once a month, and 5=at least once a week). The mean value of the three items was calculated as a measure of the COE of each respondent since the Cronbach’s α of these items was 0.898. The parents’ free time was evaluated through their average working hours (excluding housework and childcare) (Truong et al., 2023).

NEST was used to objectively evaluate the data from the six parks (Gidlow et al., 2018). This tool lists 47 items, which are divided into eight fields: Access, Recreational facilities, Amenities, Significant natural features, Aesthetics-natural features, Aesthetics-non-natural features, Incivilities, and Usability. Table 1 shows the items in each field. Most items were evaluated according to their existence and quality at the same time. Therefore, the higher the score, the more likely the item exists and the better its quality; and if the item does not exist in the park, a value of zero was assigned. Three observers evaluated each park independently. The consistency between observers was calculated, and the item scores were entered into Microsoft Excel by the observers before they were compared. The scores of items in various fields were added together, resulting in a comprehensive score for each field. The significant natural features were combined with the Aesthetics-natural features field when calculating the score since the significant natural features of urban parks are not apparent.

Only respondents whose residence was within 500 m from the park were chosen as research objects, and the normalized difference vegetation index (NDVI) to around 500 m may affect the use of the park green space by visitors (Wang et al., 2022). Therefore, the NDVI of the 500 m buffer zone of parks was taken as a control variable. NDVI is a frequently applied indicator of plant coverage in a specific area (Markevych et al., 2014). It is calculated based on the different surface reflectivity of the red (R) and infrared (IR) bands in multi-spectral raster data sets (Qin et al., 2021) and can be calculated as:

The NDVI value ranges from -1 to 1. Negative values represent clouds, water, and snow, whereas values close to zero mostly represent rocks and bare soil. Positive values indicate green vegetation, and the higher the value, the higher the density of green vegetation. In the present study, NDVI data were acquired from Sentinel-2 satellite images with a spatial resolution of 10 m taken in July 2024 (Figure 1).

First, the scores of each field were obtained by adding up the scores of the items in a field, which were scored independently by three observers. By selecting the middle of the three scores offered, any disagreements among the observers were resolved (Wang et al., 2022). Then, their intra-class correlation (ICC) was measured to test the reliability of the NEST. The ICC of the three observers was 0.975, indicating that the measurement consistency of the three observers was very high; therefore, the reliability of the NEST was also high.

Then, using 317 samples from six different urban parks, the multilevel regression model was used for investigating the factors that affect children's use of urban parks. Multilevel modeling was developed in response to the challenge of appropriately analyzing clustered data. This technique preserves the original data structure while explicitly modeling the within-group homogeneity of errors by allowing the estimation of error terms for both the individual and the group (Krull and MacKinnon, 2001). Multilevel modeling is widely used in urban management, health, education, green leisure and other fields (Roberts et al., 2019; Qin et al., 2021; Wang et al., 2022; Jun et al., 2024; Konings et al., 2024).

The multilevel regression model illustrates the shared variance of clustered data sets and can be used for resolving the clustering and non-independence problems of respondents from the same park (Roberts et al., 2019). This approach allowed us to examine the influences of parental level and park level factors on park use. Furthermore, two-level regression with random coefficients was used in the present study. This approach allowed us to examine the influences of parental level and park level factors on park use.

The Level-1 Model of the parental level was:

where Y_ij is the park use frequency of respondent i in park j, and β_0j indicates the intercept. X_1ij, X_2ij, …, X_pij indicate the parental-level factors of respondent i in park j, and β_1j, β_2j, …, β_pj represent the corresponding coefficients.

In the Level-2 Model, park-level variables W_1j, W_2j, …, W_Qj were adopted for testing the influence of park features on children's use of urban parks. As the main focus of this study was the correlation between park features and the frequency of park use by children, only the Level-1 intercept (β_0j) was allowed to change between parks including the random intercept model. Therefore, the Level-2 Model used the following format:

where γ₀₀ represents the average intercept representing the total average value of the dependent variable Y_ij. The coefficients γ₀₁, γ₀₂, …, γ_0Q represent the main influences of park-level variables, and W_1j, W_2j,…, W_Qj represent the use frequencies of the urban parks. Regarding the random effect, u_0j refers to the inter-park variability of the average intercept and it is assumed to obey a normal distribution. In addition, the “xtmixed” command in STATA16 was employed to perform multilevel regression.

Most of the respondents (60.3%) were mothers, and more than half of them had a monthly income of either 6001- 8000 or 8001-10000 yuan. More than half of them had a bachelor’s degree or above. Nearly 80% of the parents worked mainly for 8-10 hours. See Table 2.

A considerable proportion (65%) of the parents agreed with the statement of the importance of four benefits to some degree, and the average score of each item was high. The average score of health and welfare was the highest (3.88), followed by learning about nature, meaningful family activities, and establishing friendships. In contrast, the parents’ evaluation scores of accidents were low, and the average score of each item ranged from 3.20 to 2.81. The biggest concern was accidents (3.20 on average), followed by attacks by strangers (3.10 on average), and the least concern was animal attacks (2.81 on average).

In the COE of parents, only minor differences were found between various items, ranging from playing with grass, flowers, and fruits (average score of 3.48) to catching fish, frogs, or insects (average score of 3.45) to climbing trees or playing tag/hide-and-seek with a tree (average score of 3.41).See Table 3.

The respondents used Fuxing Park and Luxun Park with average frequencies of more than 10 times per month, followed by Kangjian Park and Zhongshan Park (7-9 times a month), and Yichuan Park and Jingan Park (about 6 times a month).

According to the spatial autocorrelation analysis findings, no spatial autocorrelation existed in the average use frequency of six parks (P≥0.05; Moran’s I<0), and it showed a discrete trend (Figure 2).

The park variables were objectively evaluated as variables. Because the variables from objective evaluation are highly correlated (which will eventually introduce multicollinearity problems in the model) (Subiza-Pérez et al., 2020), correlation analysis was conducted (Table 4) to select the park level variables for the multilevel models. Four features were significantly related to park use (i.e., Access, Recreational facilities, Incivilities, and Usability), which shows that park use is closely related to certain park features. To avoid multicollinearity, the variable with the highest correlation (Usability) was used as a park variable in the multilevel models (Table 5).

First, a zero model without independent variables was built (Table 5). The proposed model divided the total variance of the dependent variable (park use frequency) into the parental level and the park level. Based on the obtained results, parental level variance was 25.510, park level variance was 4.166, and ICC was 0.140, indicating that the 14% difference in children’s use of parks can be illustrated by the differences between parks (Table 5). Furthermore, the existence of a clustering effect demonstrated that the use of multilevel regression modeling was appropriate.

Then, the parental variables were introduced into Model 2. The results showed a significant negative correlation between the parents’ working hours and children's frequency of park use. The parents’ perceived benefits of park activities and their COE are positively correlated with children’s frequency of park use.

In Model 3, the park-level variable of usability was added. The results showed that higher park usability can increase the frequency of children’s use of parks.

The socio-economic characteristics of parents (including gender, income, and education level) have no significant influence on children’s use of parks in Shanghai, which differs from the research results obtained in Turkey, the USA, and other countries. For example, a US study showed that parents’ educational background has a positive effect on children’s biophilia and the importance they assign to outdoor activities (Hammond et al., 2011; Ahmetoglu, 2019). Parents with higher education levels are more likely to cultivate their children’s love for outdoor activities and help their children to interact with greenspaces. A Turkish study reported a positive correlation between family income and parents’ attention to their children’s greenspace experiences. Compared with middle-income and high-income parents, low-income parents reported a lower level of importance of interaction with greenspaces. The differences between the results of this study and those of the USA and Turkey may be due to the different social and cultural backgrounds. More research is needed to further substantiate the results of this study.

The most critical driving factor for children’s use of urban parks is the parents’ perceived benefits of their children’s activities in greenspaces, which is consistent with previous research results (Ahmetoglu, 2019). When parents realize the benefits that interaction with greenspaces has for their children, they tend to provide opportunities for them to play in the greenspaces. This indicates that the more positive parents are towards greenspaces, the more their children like to play in green spaces (Hammond et al., 2011).

In the present study, a considerable number of the surveyed parents agreed that interaction with greenspaces is good for their children, and the most apparent benefits are correlated with health and welfare, followed by learning about nature, family values, and friendship (Wang et al., 2018; Truong et al., 2022), which is similar to the findings of previous studies conducted in China and Japan. However, this result differed from a study of US parents, which reported that it is most important for children to “spend good time with their families” in outdoor entertainment activities (Larson et al., 2013). Even though the parents from these different societies and cultures hold various views on the importance of the benefits greenspaces bring to their children, they agree that contact with the greenspaces is conducive to the healthy growth and development of their children (Larson et al., 2013; Gundersen et al., 2016; Wang et al., 2018). Thus, parents in developed countries may attach great importance to their children’s contact with greenspaces. To fully comprehend the parents’ understanding of the benefits that interaction with greenspaces has for their children, it will be necessary to conduct more research in developing countries.

The present study shows that when they were children, most parents enjoyed outdoor activities once per year or once per month. The most common outdoor activities were playing with grass, flowers, and fruits, followed by catching fish, frogs, or insects, and finally climbing trees or playing tag/hide-and-seek with a tree. Thirty to forty years ago, the degree of urbanization in China was low, electronic products were not widely available, and children had more opportunities and time to engage in outdoor activities. Therefore, the current generation of parents experienced more opportunities, orientations, and higher frequencies of engagement in outdoor activities. The frequent outdoor activities during the parents’s own childhood also increased the frequency of parents taking their children to play in parks. Therefore, strengthening outdoor activities and increasing the frequency of park usage from an early age may be an effective way to promote the use of green parks in their children. However, a Japanese study found that the outdoor experiences of parents in their childhood did not increase the frequency of allowing their children to play freely in greenspaces (Truong et al., 2022). The reason is that in modern society, few parents allow their children to play in green spaces without adult supervision, which is the opposite of the freedom they had in their own childhood. Traffic safety problems and fear of strangers are the main reasons why parents cannot give their children as much freedom as they had in their own childhood.

This study also found that accidents and fear of being hurt by strangers are the most concerning accidental factors for parents (Skar et al., 2016). However, the parents’ concern for accidents was not a strong predictor of children’s use of green spaces in this study, which differed from previous studies which showed that parents’ concerns about accidents prevented their children from playing alone in greenspaces (Karsten, 2005). The reason for this difference is that previous studies focused on the influencing factors of children playing alone in green spaces, while the present study examined the influencing factors of parents taking their children to play in green parks. This shows that parental supervision and companionship can alleviate the worries that parents may have about accidents, and their companionship means that accidents are no longer a factor that limits their children’s interaction with greenspaces.

This study shows that parental working hours are a main limiting factor for children’s use of parks. More than 90% of parents work for more than eight hours. In an international metropolis like Shanghai, people experience relatively strong work and life pressures, and the parents of these children form the backbone of the workplace. The work stress and long working hours of parents limit their children’s use of parks.

This study shows that the usability of urban green space is an important factor affecting children’s use of parks. Among the six parks that were sampled in this study, Luxun Park (22.37 ha) and Fuxing Park (8.89 ha) had the highest usability scores (21 points), while Zhongshan Park (20 ha) and Kangjian Park (9.57 ha) were in the middle (12 points), and Yichuan Park (1.88 ha) and Jingan Park (3.36 ha) had the lowest usability scores (10 points). These results show that the usability of green parks has a certain relationship with the park area. The larger the park area, the higher its usability. Because large parks offer more space for activities, children can play more types of games, including ball games (Flowers et al., 2020).

Usability includes sports, informal games, walking/running, children’s play, conservation, enjoying the landscape, meeting/socializing, relaxing, cycling, water sports, and fishing. Usability mainly measures the extent to which the park is suitable for these activities. No differences were found in the usability for cycling, water sports, or fishing among parks, and the scores of all these items were 0. Cycling and fishing are not allowed in most urban parks in Shanghai, and water sports are rare. The usability for enjoying the landscape, relaxing and meeting/socializing all scored 2 points, indicating that these parks in Shanghai can basically satisfy the fundamental requirements of residents for these activities.

The six parks showed great differences in usability for sports, informal games, walking/running, children’s play, and conservation (see Figure 3), which were clearly higher in Luxun Park and Fuxing Park than in the other parks. The usability for sport, informal games, walking/running, and children’s play depends on the park space, entertainment facilities, and paths. For example, Luxun Park has spacious activity space, green space, and entertainment facilities, which are suitable for playing informal games and sports such as running, jumping, and football. Therefore, children coming to such a park will be happy and have something to do (Veitch et al., 2006). In other words, whether the park can meet the needs of children’s play, games, and other sports activities is an important factor that affects their use of green parks. Luxun Park and Fuxing Park are also higher in conservation than the other parks, which shows that good maintenance and management are important factors that can lead to usability differences between parks (Dawson et al., 2023).

Fuxing Park covers an area of 8.89 ha, which is smaller than the area covered by Zhongshan Park (20 ha). However, Fuxing Park’s usability scores in sport, informal games, walking/running, children’s play, and conservation were higher than those of Zhongshan Park. This implies that good park planning, design, management, and maintenance can compensate for the problems caused by an insufficient area to a certain extent.

The 500 m NDVI around the park was introduced into Model 3 as a control variable. The results show that the 500 m NDVI around the examined parks did not affect the use of these parks. This means that a higher NDVI around a park will not reduce children’s use of it, which further reflects the demand of children for the usability of park activity space, entertainment facilities, maintenance, and management.