Keywords

1 Introduction

1.1 Accessibility of Urban Green Space in Urban Planning

There is an increasing consensus of Urban Green Space’s (UGS) role on promoting physical activity, psychological health and social connectivity (Ghimire et al. 2017; Kaczynski et al. 2010). In China, according to the 2010 PRC National Standard, UGSs are divided into three main types: namely, Park Green Space (PGS), Green Buffer (GB) and Square Green Space (SGS). Given this standard, PGS is defined as green spaces open to the public with inherent beneficial impacts on people and environment. Numerous literatures indicate that parks offer comprehensive functions for people to engage in leisure, exercise and socialising (Zhang et al. 2013; Kaczynski and Henderson 2008; Hong et al. 2018). Specifically, greener living environments can help with reducing risk of obesity (Mears et al. 2020; Manandhar et al. 2019), relieving depression (Meyer-Grandbastie et al. 2020), improving self-esteem (White et al. 2017) and enhance social cohesion and sense of belonging (Hong et al. 2018).

A considerable number of studies have proved that residents within reasonable access to UGS are able to take advantage of the benefits (Jenks and Jones 2010). Therefore, understanding how people access urban parks plays a vital role in investigating whether the Green Infrastructure (GI) is allocated scientifically and effectively. For example, with the rapid urbanized growth in Copenhagen in the 1990s, lacking the balance between GI provision and intense land development led to the pressure on UGS network (Caspersen et al. 2006).

Planners and local government realise the importance of UGS provision. The European Environment Agency (EEA) notes that the residents should be within a 15-min walking distance from the nearest UGS. Great London Authority offers a guidance that there should be accessible UGS of at least 0.02 hectares no more than an 800-metre walk from home. Accessing UGS has been recognised as a concern for environmental justice, gaining recognition on assessing the disparity between GI provision and the public demand based on socioeconomic population groups (Xiao et al. 2017; Xu et al. 2018). For example, Boone et al. (2009) indicated that black ethnic groups have less urban park acreage within 400-m walking distance than the white ethnic groups in Baltimore, Maryland. However, there is far from a consensus on findings obtained by empirical studies. Two opposite reviews are presented by Macintyre (2007) and Wolch et al. (2014), of which the former found a better access to parks in lower-income population than other groups, whilst the latter found low-income communities are provided with less park service than white and more affluent counterparts. Therefore, the question of how to best measure UGS accessibilities has a vital influence on research findings (Lee and Hong 2013a, b; Mears and Brindley 2019; Rigolon 2016; Wolch et al. 2014).

1.2 Methods of Accessibility Modelling and Measurement

Methods of modelling the accessibility of urban parks still remain debated within urban planning and the sustainability of cities (Dadashpoor and Rostami, 2017; You 2016; Xing et al. 2018). Geurs and Van Wee (2004) discussed that the accessibility measurements can be divided into four perspectives, including infrastructure-based measure, individual-based measure, location-based measure and utility-based measure. Park accessibility measures are seen as indicators for the effect of distribution and quality of parks and transport systems on people’s access to parks (Xu et al. 2017). Based on this concept, and referring to Geurs and Van Wee (2004), the constitution of park accessibility measures can be regarded as coming in four parts: namely, park component, transportation component, temporal component and individual component. The interrelationship among the four components is depicted in Fig. 13.1.

Fig. 13.1
A schematic. Accessibility to opportunities has 4 interrelated components namely land-use, transport, temporal, and individual, with their features.

(Adopted from Geurs and Van Wee 2004)

Interrelationship among four components of accessibility measure.

Full size image

Following this framework of measuring accessibility, Table 13.1 lists some measuring and modelling standards and methods of UGS accessibility from the literature. Common measurements calculate the amount of UGS per capita or add up the average/total time/distance to access to UGS (de Sousa Silva et al. 2018; Rigolon 2017; Quatrini et al. 2019). The travel time or distance can be estimated by a simulation model (Kim et al. 2018) such as the buffer area (Tratalos et al. 2007; de Sousa Silva et al. 2018), network analysis (Rigolon 2017; Quatrini et al. 2019), collected from floating car data (Nigro et al. 2018), or captured via an online map API (Chen et al. 2017). Their main drawback is failing to consider the potential impacts of other components and the interactions among them.

Table 13.1 Some standards and methods of modelling and measuring of UGS accessibility according to the literature
Full size table

To fill the research gap of a people-oriented accessibility measurement, a considerable number of studies have focused on people’s activities to investigate urban park accessibility by using such methods as collecting activity diaries (Chai et al. 2009), GPS techniques (Chai et al. 2013; Vich et al. 2019), online POI data or mobile phone signalling data (Li et al. 2011; Liu et al. 2012; Zhai et al., 2018; Mears et al. 2021). Such studies, however, frequently overlook the effects of the ‘supply’ component, and it is difficult to interpret the characteristics of individuals.

Other research employs a gravity concept for modelling the UGS accessibility, which considers the way that attraction of destination affects the accessibility and combines the influences of land use and transportation. However, the interaction between supply and demand is not reflected. To improve the original gravity-based model, Luo and Wang (2003) firstly posed the two-step floating catchment area (2SFCA) method, which subsequently has become widely used in the field of analysing access to health care (see Chen et al. 2021; Ghorbanzadeh et al. 2021; Kiani et al. 2021). During recent years, researchers have applied the method to evaluating GI accessibility (Lin et al. 2021; Liu et al, 2021; Xing et al. 2020). However, the content of data does not embrace personal and preference information, potentially overlooking the relationship between public behaviour and public perception on using parks. For example, when some studies use a gravity model, most of them regard only the size of green space area as assessing the factor of GI’s attractiveness indicator.

As discussed above, deficiencies in current UGS accessibility studies remain, including indicators of park accessibility that lack an investigation into whether there is a supply–demand imbalance of GI allocation. It has been demonstrated that people do not always visit the nearest parks given the quality of destinations (Vaughan et al. 2013) and different transportation modes (McKenzie 2014). Different population groups who walk to access UGS are demonstrated to be closely related with significant social inequality in Shenzhen, China (Xu et al. 2017). Additionally, walking is the most equitable mode of transport, available to everyone (Arellana et al. 2021). Walking accessibility of urban parks is crucial to assess the rationality of GI provision under the context of urban planning and environmental justice.

Given the background and gaps mentioned previously, this paper aims to answer the following questions: (1) What is the spatial walking accessibility of urban parks in Guangzhou? What is the actual walking accessibility perceived by the residents? (2) To what extent is there a disparity between modelled accessibility and public perceived accessibility of urban parks? Where are the regions that lack a sense of proximity to UGS? (3) What factors could potentially lead to such disparities in UGS accessibility? More importantly, this empirical study contributes to improving traditional UGS accessibility modelling and provides insight into urban parks planning for local administrators from the residents’ perspectives rather than a traditional top-down mode.

2 Materials and Methods

2.1 Study Area

This empirical study is conducted in the southern Chinese city of Guangzhou, the capital of Guangdong Province and a major area of the Pearl River Delta. It lies adjacent to Macao and Hong Kong and covers an area of 7434 km2. It is currently one of the most populous cities in the mainland of China with a permanent population of 18.6 million in 2020, which has increased 64% in the past 13 years. According to the revised administrative division of Guangzhou in February 2014, it consists of six main districts and five satellite districts (Fig. 13.2). Our research focuses on the central-city area that is constituted by its six districts: Yuexiu district, Haizhu district, Tianhe district, Liwan district, Baiyun districts and the new Huangpu district (including the old Huangpu district and Luogang district) as shown in Fig. 13.2a. In this region, less than 10% of population is employed in agricultural work, such as farming and fishing (Chen and Yeh 2018).

Fig. 13.2
3 maps. a. A map of China locates Guangdong and its urban area at the center of the province. b and c. A map of urban Guangdong has a high population distribution and the majority of parks with numerous entrances in the southwest end.figure 2

Images Yueshan Ma

Location of study area in China and the administrative division of Guangzhou a Study area is in the central region of Guangzhou indicated in orange; b Population distribution; c Parks distribution and entrances.

Full size image

Local government has facilitated the construction and protection of GI under the intense state of land sources. The Forestry and Garden Bureau (FGB) of Guangzhou municipality (2018) stated that, by 2035, the per capita green space area of parks should be raised to no less than five square metres and citizens should be able to visit the nearest parks within a ten-minute walking time. Additionally, it aims to make the public the beneficiaries and supervisors of park green spaces. By the end of 2035, there will be more than 800 new parks in Guangzhou. The central city area takes up around 20.3% of the land area of Guangzhou city area and has about 52.8% of the park green space area of the whole city. All the parks with free admission located in the central city area are included in the study. In addition, parks within the 500-m buffer area surrounding the central city area are selected as study objects for avoiding ‘boundary effects’.

2.2 Data Source and Pre-processing

The following data were utilised:

  • Green spaces and road networks from OpenStreetMap (OSM);

  • Age and other population demographics from the census alongside administrative boundaries from the Statistic Bureau of Guangzhou Municipality;

  • Population counts at 250-m grid cells from Global Human Settlement Layer;

  • Additional green spaces not found in OSM from the Bureau of Forestry and Landscaping of Guangzhou Municipality.

It is possible to simply aggregate the fine spatial population data from the Global Human Settlement (250-m grid cells (Schiavina et al. 2019)) to the much larger areal units associated with township data from the Statistic Bureau of Guangzhou Municipality (average township area is around 1,267 ha). Therefore, the population grid data (Fig. 13.2b) was spatially joined to the townships, and census data was added to each town based on corresponding town names. Guangzhou encompasses 170 townships in which the central city districts include 116 townships under jurisdiction of six districts. As stated before, the UGS polygons were filtrated by the type of Park Green Space (PGS) and only parks located in the central city area and its 500-m buffer were chosen. In total, 237 parks were included (Fig. 13.2c), which vary in function, location and size. Entrances of parks were extracted from OSM, but for those parks without an entrance the geometric centroids were utilised. In reality, it should be noted that there is no reliable open source of entrances for the hundreds of parks in the city.

2.3 Methods

This study utilises the Two Step Floating Catchment Area (2SFCA) method proposed by Luo and Wang (2003). This measurement quantifies the relationship between supply and demand. The first step is generating a service catchment (Sj) with the travelling cost (djk) for each destination (j), and then adding up the population (Pk) within this area to obtain an area-to-population ratio (Rj). The second step is to accumulate the area-to-population ratio (Rj), in which the population contains people within the catchment (Pi) covering a distance (dij) from each population location (i), and the area (Ai) is the sum of parks located in the population catchment. The distance is set as a constant of 1000 m according to the public preferred walking time of 15 min (the preferred walking time results from the questionnaire output–and is also supported by literature, such as the European Environment Agency guidance).

Step 1:

For each  service  (j), \(R_{j} = S_{j} /\sum\nolimits_{{k \in \left\{ {djk} \right.}} {P_{k} }\).

Step 2:

For each  population  (i), \(A_{i} = \sum\nolimits_{{j \in \left\{ {dij} \right.}} {R_{j} }\).

The primary data collection method involved both an online and on-site questionnaire which employed a random sampling strategy to collect 2505 responses (2375 online and 130 on-site). All participants were aged 18-years or over and stay or have stayed in Guangzhou more than three months. This field survey was conducted from October 2019 to January 2020, and the online survey was continued from November 2019 to November 2020. As such, the period covered all seasons of the year. These random samples were tested to be representative of the actual population characteristics of Guangzhou (including age groups, gender groups and areal distribution) and were also normally distributed.

The 2SFCA method and service area analysis are mainly used for modelling the spatial accessibility of parks. The questionnaire survey contributed to both qualitative analysis and quantitative analysis including statistical analysis and analytic hierarchy process (AHP). The modelling process used ArcGIS, and the statistical analysis was undertaken in SPSS. It should be noted that AHP is utilised as a tool for obtaining the coefficient of factors for representing the accessibility of parks. The factors are selected based on previous studies, and the priority of factors is decided by answers from the questionnaires (https://www.wenjuan.com/s/2iQzAvX/).

3 Results

3.1 Two Types of Accessibility

We analyse the accessibility of parks in the central city area using two measures. The first describes the modelled spatial relationship between park supply and population demand using the 2SFCA method. The final accessibility values are reclassified into five levels by geometric interval classification and allocated to corresponding townships. The second measure represents the public perception on access to parks. The grades are rated 1–5 by respondents on the basis of their ease of walking to parks (1: Very hard; 2: Hard; 3: Medium; 4: Easy; 5: Very easy). The statistical description of accessibility levels is shown in Table 13.2. This shows that in aggregated terms, the accessibility of parks based on residents’ actual use (PPAG) appears broadly comparable with those from the spatial modelling (MSAG). To some extent, the PPAG is slightly higher than the MSAG in total.

Table 13.2 Descriptive statistics on overall accessibility (both modelling accessibility and perceived accessibility) in central city area
Full size table

The distribution of classified modelling accessibility levels is shown as grades in Fig. 13.3. It reveals potential deficiencies in park supply in the south-west region of the central city area, where although the park entrances are greater in number, the population is more densely distributed than in other areas (Fig. 13.2b). Among regions with modelling spatial accessibility grades (MSAG) less than three, the lowest MSAGs are found in townships in Liwan district and Yuexiu district, taking up 26.3% equally. Huangpu district has the least proportion of low MSAGs with 5.3%. Among townships with a high MSAG (>3), the most belong to Tianhe district with the proportion of 29.6%.

Fig. 13.3
A map of urban Guangdong. Park entrances have a high density distribution in the south. The spatial accessibility regions are, 1. southwest, very low to low, 2. southeast, medium to high, 3. north, very high.

Image Yueshan Ma

Modelling spatial accessibility grades (MSAG) of Parks at township level.

Full size image

The public-perceived accessibility grade (PPAG) is obtained from people’s self-rated satisfaction on access to parks in their townships as presented in Fig. 13.4a. The low PPAGs (<3) are mostly found in the Yuexiu district (26.9%) and the Tianhe district (23.1%) and rarely appear in the Huangpu district (3.8%), which matches with MSAG distribution discussed above. In terms of high PPAGs (>3), Baiyun district includes the highest PPAGs accounting for 24.4%. However, when comparing the PPAG with its corresponding overall experience on visiting walkable parks (Fig. 13.4c), many high-MSAG regions exhibit lower satisfaction. In contrast, in some low PPAG regions, although people feel that there is not enough access to parks, visiting these parks could fulfil their requirements for UGS within a walkable distance. To some extent, this corroborates the findings of previous studies that people do not always visit their nearest parks. In other words, spatial distance should not be the only factor in representing the actual walking accessibility of parks based on public usage.

Fig. 13.4
2 maps of urban Guangdong. a. The number of parks and their entrances decrease while the accessibility increases from south to north. b. Public satisfaction increases from south to north.

Image Yueshan Ma

Perceived accessibility levels of parks a and corresponding satisfaction regarding visiting walkable parks b

Full size image

3.2 Public Perception on Visiting Parks

After exploring the relationship of user ages and users’ preferred resident-reported walking times and acceptable longest walking times to parks, Table 13.3 shows a significant correlation between walking time and age. This demonstrates that up to the age of 50, people are willing to take a longer time to walk to visit a park (Table 13.4). The time range of ‘2’ represents 10 to 20 min and ‘3’ relates to 21 to 30 min. There is a decrease in preferred walking time for people over 50 years old.

Table 13.3 Correlations between age and walking time with significant relationship (Sig. value = 0.00)
Full size table
Table 13.4 Descriptive statistics for the relationship between walking time (original time range is from 1 to 5, where 1–‘less than 10 min’; 2–‘10–20 min’; 3–‘21–30 min’; 4–‘3–40 min’; 5–‘more than 40 min’) and age
Full size table

As discussed above, the actual public perception on access to parks may not reflect the modelled accessibility. Therefore, in addition to travelling costs (expressed as travelling distance) there must be other factors affecting people walking to parks. Factors are identified from analysing their importance within the questionnaire responses, including: park area, facilities (toilets, fitness equipment etc.), views (views within the park), cleaning (clean environment in parks), travel time, reputation (perceived park popularity), activity diversity (public benefit activities, sports activities etc.), traffic safety (on the route to parks), environmental quietness (noise in the park from surrounding traffic, crowd noise etc.), surrounding infrastructure (markets, restaurants, etc.) and surrounding security (when walking in and nearby the park). The public scored views as the most important factor influencing how far they would be willing to walk to parks (Fig. 13.5). The joint second-most important factors were cleanness, travel cost and level of facilities, with the park area being the next ranked factor (Fig. 13.5). The influence of the surrounding security and infrastructure were comparatively lower rated factors. Additionally, these results offer the specific coefficients of these factors after conducting the AHP (Table 13.5), which can be used as reference for modelling actual accessibility from user-based aspects.

Fig. 13.5
A vertical bar graph of number of respondents versus factors. View, 1400. Cleaning, travel time, and facility, 780. Park area, 700. Diversity of activities, 500. Reputation, 420. Environmental quietness, 400. Traffic safety, 350. Surrounding security and infrastructure, 250. Data are estimated.

Factors affecting access to parks on foot (y-axis represents the number of respondents selecting this factor)

Full size image
Table 13.5 Weights of factors in Fig. 13.5 affecting access to parks on foot
Full size table

4 Conclusion

The results of our case study show that, in total, the accessibility level is average in the central city area. Specifically, Yuexiu district and Liwan district recorded MSAG values of 2.5, which is lower than the average level of 3. Contrarily, Huangpu district shows the best access to parks with a MSAG of 4. Discrepancy between modelling and actual accessibility is detected via comparing MSAG and PPAG within the same region. From the public perception, the MSAGs of Yuexiu district and Liwan district rank equally last with scores of 3.5. Baiyun district and Huangpu district rank equally first with MSAG of 3.9. When adding geographical distribution characteristics to these grades, there are several townships having high MSAG but low PPAG, including three townships in Yuexiu district, five districts in Haizhu district, three townships in Liwan district and five districts in Huangpu district. This difference might indicate that planners should pay more attention to other factors in addition to distance in order to improve the accessibility of parks, especially in these regions. In this respect, this case study has provided suggestions on weighting indicators that influence people’s walking to parks. The view of parks was the most important factor and considered more important than travel cost which ranked second.

As an empirical study, these results can also help to inform park green space planners and policy makers in Guangzhou about both spatial provision of parks and public perception of park accessibility. The research highlights the novel approach of comparing modelled and user-based accessibility to explore park accessibility. Discrepancies between the two measures serve as a reminder of the additional hidden factors that significantly affect accessibility of park green spaces.