1 INTRODUCTION

Public open green space is essential for healthy, liveable, and sustainable urban environments and a priority under the United Nations (2016) Sustainable Development Goals. Urban green spaces fulfil numerous biophysical, social, and cultural purposes (Grose, 2009) to manage urban water, protect biodiversity, and reduce urban heat island effects. Urban green spaces also provide settings or facilities for physical activity; promote mental well-being and support attention restoration, stress reduction, and positive emotions; and foster social well-being through social interaction and participation (Bowler et al., 2010; Dinnie et al., 2013; Koohsari et al., 2015).

In geographical, public health, and physical activity research, “parks” have typically formed the basis for exploring urban green space and health outcomes (Hooper et al., 2020). Researchers have often taken a “recreational opportunity spectrum” approach (Clark & Stankey, 1980) focused on physical and spatial aspects of park provision and measures of the availability and accessibility to parks (Bancroft et al., 2015; Lamb et al., 2019) or of the “quality” of parks (Bedimo-Rung et al., 2005; Zhang et al., 2018). These methodological approaches typically use buffers around participants’ homes or compute distances to the closest park from homes and assess associations between parks and health outcomes (Lamb et al., 2019). However, such an approach assumes that participants only use the park closest to them in a defined residential catchment area, even though other parks are available to visit outside designated neighbourhood buffer areas (Kaczynski et al., 2014). Few studies identify the exact parks that participants use, although use is often hypothesised in the access–use–health relationships being tested. Indeed, Edwards et al. (2015) have found that only 27% of adolescents used the park closest to their residence.

Internationally, planning for parks or urban green spaces is typically guided by a standards-based approach that stipulates targets for the provision of a minimal quantity of green space per capita or the maximum distances residents should have to travel to access parks (Hooper et al., 2018). These standards help ensure equitable distribution of parks, including desired facilities, amenities, and programming across a community, but they neglect to consider how members of communities use the spaces (Veal, 2013).

Measures of visitation to parks are important for understanding patterns of green space use in urban settings. Focused on the impact of public resources, governments at all levels are challenged when it comes to showing the impact or benefit of these recreational spaces (Spangler & Caldwell, 2007). Hence, the evaluation of public spaces is becoming increasingly important as population densities increase and competing interests for land allocation arise (Tyrväinen et al., 2007). Essential, but currently lacking is a better understanding of factors that influence the use, preferences, experiences, and values of urban parks and the benefits gained by members of the community from using them (Dinnie et al., 2013; Veitch et al., 2021). Addressing those gaps will foster more strategic, evidence-based decision-making related to future park design, service provision, maintenance, or investments to accommodate community needs. Such refinements to decision-making should enable local governments to balance competing demands on land allocation and assets or resources (Dinnie et al., 2013). Additionally, to advocate effectively for improved park infrastructure, public health and allied professionals with strong spatial skills must understand and demonstrate the benefits that parks have for health behaviours and outcomes (Henderson & Fry, 2011).

Local government community surveys typically collect information on the demand for participation in outdoor recreation, park use, and attitudes. In Australia, community surveys are often ad hoc and primarily used when a new public open space strategy is being developed, which may be every 5 to 10 years. Moreover, the type and quality of the data typically vary considerably across local governments, with little or no consistency in survey items across jurisdictions.

The accurate measurement of park use is a missing link between understanding interactions among people, urban green space environments, and health benefits, and it requires further exploration (Chen et al., 2020). This knowledge gap hinders new approaches to park transformation and renovation, urban planning and park design, and the development of environmental change interventions to improve population physical activity and public health. However, the measurement and surveillance of park use require appropriate, reliable, and accurate instruments.

In response to these critical research gaps, Section 2 provides a review of different methodological approaches that have been used to measure park use in the geographical and public health fields. Section 3 presents work on existing public participatory geographic information system (PPGIS) instruments that have been developed to assess the use of parks or natural environments. Section 4 advances the field of PPGIS and presents a test–retest reliability study of the Park Life PPGIS instrument for the identification of parks used and the activities undertaken in parks.

2 APPROACHES TO MEASURING PARK USE

In population health and physical activity literatures, several approaches have been employed to measure park use (Joseph & Maddock, 2016; Sylvia et al., 2014). Numerous studies have employed observation approaches and momentary time sampling methods to record park users behaviours using tools such as the System for Observing Play and Recreation in Communities (SOPARC) (Cohen et al., 2011; McKenzie et al., 2006) and the Environmental Assessment of Public Recreation Spaces (EAPRS) (Saelens et al., 2006). A few studies have used uncrewed aerial vehicles such as drones and wearable video devices to record park use and park-based physical activity (Park & Ewing, 2017; Suminski et al., 2021). Data are then retrospectively analysed to count park users and code activities. A limitation of such approaches is that each observational protocol only tells researchers about the use of the parks being studied or observed. Thus, only a few parks within given regions are assessed, making it impossible to accurately generalise findings to determine trends in these regions and across a community or in relation to the multiple parks people may use (Joseph & Maddock, 2016). Indeed, spatially explicit, perceptual, and use information at city-wide scales is scarce (Rall et al., 2019).

Global Positioning System (GPS) technology has also been used to identify the parks that are used (Evenson, Wen, Hillier, & Cohen, 2013; Maddison & Mhurchu, 2009; Stewart et al., 2016). However, despite the increasing reach and capabilities of smartphone technology, which could be used for research purposes, the participant burden and privacy concerns associated with location-based research may explain the small number of studies using the technology to date (Evenson, Wen, Hillier, & Cohen, 2013; Stewart et al., 2016). Moreover, GPS provides no information about the reasons or motivations for parks used or the types of activities undertaken in them (Chen et al., 2020).

Other methods for analysing park use involve “big data” from location-based social media networks (Martí et al., 2019). User-generated location-based data are referred to as volunteered geographic information because the expressed perceptions, interests, needs, and behaviours are published online, voluntarily, by users (Martí et al., 2019). For example, some studies have examined park use based on the count and visual content of photographs geolocated within parks (Donahue et al., 2018; Hamstead et al., 2018; Sessions et al., 2016; Song et al., 2020). Others have analysed the language of social media posts (Instagram), reviews posted of urban parks on Google (Niță et al., 2021; Park et al., 2022), and Twitter [X] posts relevant to park and nature experiences and recreational activities (Park et al., 2022).

However, these data are not without limitation. For example, concerns have been raised about lack of representativeness of both participants and parks. The use of social media platforms varies according to age group and willingness to geotag posts or share content openly (Arribas-Bel, 2014; Heikinheimo et al., 2017; Heikinheimo et al., 2020; Martí et al., 2019). Only a small portion of Twitter users activate the geocoded function when publishing tweets (Martí et al., 2019), and data retrieved from Instagram are not georeferenced to the exact location from which they were posted. Instagram has delimited areas with a geolocated centre point to which all data within the area will be associated (Martí et al., 2019); thus, identification of any park used is not reliable. Additionally, because no personal details are retrieved when extracting data, a sample cannot be rigorously characterised in terms of user profiles or demographic characteristics (Chorley et al., 2015; Martí et al., 2019). While automated content analysis of social media data has been used to gain insights about revealed activities, interpreting photograph content may be less straightforward because the perception and meaning of what a photograph depicts can be highly subjective (Oteros-Rozas et al., 2018; Song et al., 2020). Moreover, reported “park use” behaviour may depend on the context of a park visit, which has implications for park use data that represent typical park use. For example, users may be less likely to tag themselves or share a photo in a typical neighbourhood park regularly used compared with a visit to a larger, unique, or destination park (Zhang & Zhou, 2018). Indeed, validation against nationwide household surveys has shown that park photography reflects residents’ favourite parks better than their self-reported frequency of visits (Song et al., 2020). Moreover, the data provide no information on where the user has come from, how often they use the park, or the activities they engage in at the park.

As such, self-report survey measures of park use remain widely used for larger scale park use studies and are crucial for understanding people’s values and motivations behind outdoor recreation (Tieskens et al., 2018). In a small number of recent studies from Australia (Edwards et al., 2015; Hooper et al., 2020), Singapore (Petrunoff et al., 2021), and the United States (Walker et al., 2009), participants were asked to describe the parks they used, which were subsequently mapped. However, in each case, the mapping required that the reported parks be matched with those in spatial databases of parks. This task is time-consuming, particularly if there are discrepancies between park names reported by participants and official park names. Moreover, people’s perceptions or uses of parks do not always align with the demarcated areas identified or zoned and managed by local governments as public open spaces. As a result, use patterns and attitudes towards informal or unmanaged components of urban green space are typically not recognised or included in studies of park use.

An alternative approach for capturing park use data that overcomes many of these methodological shortcomings is online-based PPGIS. The term PPGIS refers to diverse methods that engage people in generating spatially explicit information for various planning and decision-making purposes (Tulloch, 2008). Mapping activities are common urban planning community engagement strategies (Kahila-Tani et al., 2019). With the development of spatial web technologies, online geo-questionnaires and PPGIS are considered effective tools to deepen understandings of quantitative data (Saadallah, 2020) and to use in urban and regional geography and planning practices (Kahila-Tani et al., 2019). PPGIS surveys can measure uses, experiences, and perceptions in various social contexts (Dinnie et al., 2013). Participants typically identify spatial locations on a map and provide data specific to that location for analysis. Importantly, PPGIS maps can be set to collect data at varied scales and enable respondents to focus on their chosen area. Thus, researchers can explore the health outcomes associated with reported place data without enforcing preconceived neighbourhood boundaries. Another advantage of PPGIS is that it reduces data collection costs and increases the precision and efficiency of spatial data entry and collation (Brown & Reed, 2009). In the context of studying park use, PPGIS generates or captures high-quality data so researchers and practitioners can understand who is using parks, what parks they are using, and why they are using these spaces (Heikinheimo et al., 2020).

We identified 13 studies that used PPGIS methods that focused on urban green spaces (Bijker & Sijtsma, 2017; Brown et al., 2014, 2018; de Vries et al., 2013; Heikinheimo et al., 2020; Hughey et al., 2021; Korpilo et al., 2021; Latinopoulos, 2022; Pietrzyk-Kaszyńska et al., 2017; Rall et al., 2019; Ramírez Aranda et al., 2021; Raymond et al., 2016; Schrammeijer et al., 2022). Nine studies explicitly asked participants to identify and map the green spaces they used. However, the specific instructions and recall periods differed across all studies (for example, “visit most” [Schrammeijer et al., 2022], “spend the most time in” [Latinopoulos, 2022], “used in the past 30 days” [Hughey et al., 2021], “frequently visit” [Korpilo et al., 2021], “used for activities in the past two weeks,” “places where you spend time within green spaces” [Pietrzyk-Kaszyńska et al., 2017], and “areas you enjoy using” [Raymond et al., 2016]). The remaining four studies were less explicit in their instructions, asking participants to map green spaces that were their “favourite” (Raymond et al., 2016), “meaningful” (Rall et al., 2019), or “attractive, valuable or important” (Bijker & Sijtsma, 2017; de Vries et al., 2013).

However, people may value parks, green spaces, or natural environments without visiting or using them. There was also some ambiguity about what is measured in these studies. Phrases about the “activities” engaged in within the spaces, the “motivations” for use, or reasons the parks were “attractive” for use also differed across the studies from asking about “activities” (Brown et al., 2014, 2018; Raymond et al., 2016), “benefits” (Brown et al., 2014, 2018), “cultural practices” (Ramírez Aranda et al., 2021), “cultural ecosystem services” (Rall et al., 2019), or “qualities” (Bijker & Sijtsma, 2017; de Vries et al., 2013). These differences make comparison across studies difficult. Except for a recent study by Hughey et al. (2021), the use of PPGIS to measure park use and physical activity has, to date, not been deployed in public health-focused research.

Furthermore, participatory mapping relies on individuals’ abilities to recall their experiences in places or landscapes and locate the potential affordances of these spaces on a digital map (Brown & Kyttä, 2014). It is essential to understand whether people can reliably recall details of their park use and associated activities and confirm if PPGIS instruments are suitable for collecting these data in population monitoring and surveillance studies of park use and health outcomes. To date, studies testing the validity and reliability of PPGIS technologies and methods, including those to capture park use, have not been conducted (Chen et al., 2020). This lack of reliability data remains a significant gap in the field and applications of PPGIS.

3 METHODS

Following ethics approvals, the “Park Life” PPGIS was developed as an online geo-questionnaire in which participants individually answered a series of questions that were accompanied by interactive maps that provided geographical context about their park use. There were no constraints on the location or spatial extent of parks that could be identified or on the number of parks that could be mapped. The spatial capture of data allowed the research team to easily integrate in geographic information system, the spatially linked park data, and individual participant information and characteristics. The survey was built to capture data Australia-wide, but for this test–retest study, data collection was focused on Perth, the capital city of Western Australia. The survey questions and protocol are outlined in Figure 1. All questions were informed by previously implemented local government surveys, park literature, and a panel of experts comprising local government parks and planning managers, urban planning officers, and planning consultants.

Social demographic variables often moderate the built environment and human health relationships. As such, demographic data typically used for studies that examine associations between the use of green space and health outcomes were collected. Demographic questions captured information on participants’ gender, age, education, income, employment situation, living arrangement, dog ownership, and sports club/team membership.

Participants were recruited by a market research company that maintains a database of panellists and randomises participants to specific surveys depending on the demographic requirements of the study and that provides participants with modest compensation. Panellists who lived in the Perth metropolitan area in Western Australia were invited to participate. To assess test–retest reliability, the Park Life PPGIS participants (n = 150) were sent the survey link for a second time 1 week after completion of the first survey. On both occasions, participants were emailed a link to the Park Life PPGIS, which was conducted online.

All analyses were performed using SPSS Version 26.0. Kappa scores (κ) were used to assess the reliability of the categorical survey items. Intra-class correlations (ICCs, two-way mixed-effects model with 95% confidence intervals) were used to assess the reliability of continuous survey items. Interpretation of the coefficients relied on agreement-level ratings as suggested by Landis and Koch (1977): 0–0.2 = poor agreement; 0.21–0.40 = fair agreement; 0.41–0.60 = moderate/acceptable agreement; 0.61–0.80 = substantial agreement; and 0.81–1.0 = near-perfect to perfect agreement.

4 RESULTS

Ninety-one participants completed the Park Life PPGIS twice. The demographic characteristics of the study sample are presented using data from Test 1 and are outlined in Table 1. Sixty-two per cent of the study participants were female, 50% were in full or part-time employment, 28% had children living at home, almost 30% had a dog, 27% were a sports club member, and the majority (82%) were born in Australia.

5 DISCUSSION

Those in public health and parks and recreational planning are invested in promoting park use. Public health advocates—including urban geographers and planners—have a stake in fostering people’s visits to parks to promote health-enhancing behaviours such as physical activity. Geographers and planners have a stake in fostering visits as part of their mission to study and/or provide quality recreational and leisure experiences (Kruger et al., 2007). As such, the measurement of context-specific behaviours is vital to improving the predictive capacity for studies that measure environmental correlates of health behaviours and outcomes. However, a better understanding of the community’s use of parks is required, which in turn necessitates the need for reliable instruments to measure park use. PPGIS has grown in popularity as a tool to capture spatially contextualised data for use in urban and regional planning practice and academic research in planning-related and built environment fields (Kahila-Tani et al., 2019).

Published evaluations of the effectiveness of PPGIS have focused on participation rates and spatial data quality and accuracy (Kahila-Tani et al., 2019). In this paper, we have examined the test–retest reliability of the Park Life PPGIS for capturing park use and activities. To our knowledge, this study is the first to undertake a test–retest reliability analysis on a PPGIS designed to capture urban park use. The test–retest results indicate that individuals were able to reliably recall the parks they use most often versus those they use regularly and could map these spaces on a digital map.

The Park Life PPGIS assesses “usual” park use behaviour rather than a defined recall period such as the last week or 7 days. The repeatability of the “usual park use” question was substantial, indicating that participants were able to understand the construct of the question and reliably indicate their usual level of park use. The use of a “usual frame” in physical activity measurement questionnaires (Giles-Corti et al., 2006) has been found to have minimal differences in psychometric quality compared with the past week’s questions and minimise inherent weekly variation in physical activity (Doma et al., 2017). Conversely, a previous park use study found higher reliability coefficients for park visit frequency and duration of use with a “usual week” frame than a “past week” recall (Evenson, Wen, Golinelli, et al., 2013). A usual structure also provides a more stable measure for use in longitudinal or experimental studies seeking to monitor changes in park use behaviour in the same individual over time (Giles-Corti et al., 2006). This detail is also important for detecting a change in park use for monitoring or surveillance applications.

Recall and mapping of the park used most often, and additional parks used regularly, was very high, with kappa coefficient values in the near-perfect range. Importantly, this suggests that participants could reliably differentiate between parks used most often and those parks used regularly. Furthermore, participants were able to locate their parks on the digital map. The use of Google® base maps was chosen to increase the familiarity with the user experience when mapping the parks. A recurring question for those using PPGIS as a research method or data collection instrument has been the spatial accuracy of the data collected with a presumption of PPGIS marker placement validity (Brown et al., 2015). In this study, spatial accuracy was less of a concern, given that markers were intended to be associated with a park. The potential for spatial error was also managed by displaying polygons representing park locations in the map window to help guide the participant. The Google® search bar also allowed participants to search quickly for the park they wanted to locate. Moreover, to improve spatial accuracy, participants who selected a large park (>10 ha) were asked to move the pin to the area of the park they typically used.

The parks people use and the activities they undertake are shaped by a complex interaction between their individual needs, preferences for characteristics of specific locations, and perceived accessibility and suitability of the park locations (Chen et al., 2019). Considering a range of different recreational activities undertaken in parks is therefore essential to understand the various benefits provided by urban parks (Chen et al., 2019; Krellenberg et al., 2021), better inform planning decisions and public health interventions, and facilitate the development of liveable urban areas that promote park use, and health and well-being benefits (Hansen et al., 2019). All park activity items showed moderate/acceptable or higher reliability, with five activities showing substantial or near-perfect agreement, indicating that people could reliably recall their park activities. The high kappa values of the activities reported in multiple parks used indicate that participants were able to reliably and accurately recall different activities undertaken in multiple parks.

The detailed data elicited from PPGIS situate the Park Life PPGIS as a viable method that could be used to create park “quality” scores based on reported amenities used in parks, rather than simply assigning scores based on the number of features or amenities present. Previous park quality scores have relied on objective data primarily related to facility provision, which fails to account for subjective elements of quality or ecological, cultural, spiritual, or social valuation of the spaces (De Vreese et al., 2016). Capturing park users’ activities and experiences allows for a more balanced approach to parks planning (Rall et al., 2019). The list of park activities used in this study is applicable to other parks in Australia. However, when applied in other countries or settings, the activities may require modification.

Two distinctly different mapping approaches have been employed in earlier applications of PPGIS of urban green space use. The first approach asked participants to identify and map green spaces used or valued and then elicited information on activities, values, or motivations for use associated with that green space, from either a list of options or open-ended responses. The second approach provided a series of pins representing different activities or values that participants marked on the map to define the green spaces in which they engaged in those respective activities or were associated with the individual values. The Park Life PPGIS employed the first method, because the alternative approach results in a multiplicative effect of the response burden with the placement of each marker (Brown et al., 2015). Moreover, identifying a behavioural context or setting may assist with the recall, which may account for the high kappa coefficients.

The mapping approach used in the study presented several advantages over the use of social media generated data opportunistically extracted by researchers. For example, Park Life PPGIS pinpoints the parks and open spaces used, avoiding the pitfalls associated with the spatial inaccuracy of photo-sharing platforms (Muñoz et al., 2020). The Park Life PPGIS approach enables green space planners to consider the quantity and distribution of parks where people can pursue various activities, as well as meet the needs and expectations of different kinds of users (Bertram & Rehdanz, 2015). This approach may also include assessing the values and uses of informal green spaces, which often do not receive attention in green space planning (Rall et al., 2019).

Although this study makes a unique contribution to the evidence, it has some limitations. First, the participants were recruited through a market research company and financially compensated for their participation, which may have implications for the accurate depiction of park use behaviour. By using an external market research company, we do not have information of the demographic characteristics of participants who did not complete both the test and retest surveys. However, the sample’s demographic characteristics indicated demographic diversity in the population and the sample size in this study is comparable with other physical activity test–retest studies (Brown et al., 2004; Frehlich et al., 2018; Giles-Corti et al., 2006). Second, just under a quarter of participants (24.2%) reported and mapped using two parks. Even fewer (5.5%) identified three parks they regularly used. However, among these participants, there was no attrition between the test and retest survey in the number of parks mapped, suggesting that there were no problems associated with user willingness or acceptance to map these. Third, there was limited variability in the activities listed for the second mapped park regularly used. This fact is likely due to the small number of participants identifying a second park. Again, the kappa coefficients for these activities showed “moderate/acceptable” or higher reliability. The slightly lower kappa coefficients for the second park activities (compared with the park used most often) might indicate greater variability in the activities undertaken in these parks and/or the less regular use.

The use of PPGIS to capture park use could be easily integrated into local government public open space community consultation surveys to provide benchmarks of community park use and benefits and evaluate any park upgrades or interventions. It should be coupled with instruments to quantify the contributions of park-based physical activity to total physical activity levels, such as the Park Physical Activity Questionnaire (Park-PAQ) that has been shown to have acceptable reliability in providing a total measure of physical activity and determining the proportion of an individual’s total physical activity undertaken within parks (Edwards & Hooper, 2023).

In summary, the Park Life PPGIS was found to be a reliable instrument to capture park use and activities. The Park Life PPGIS advances the measurement of park use behaviours by explicitly addressing the missing “exact park location” data in public health research and public open space planning policy and practice. This study confirms that people can reliably recall details of their park use and associated activities and locate the spaces they use on a digital map. Further, the collation of exact park location data, coupled with park-based activity data, provides accurate data needed to enable better public open space planning.

Moreover, this reliability study advances the field of public participatory mapping methods and provides evidence on the reliability of PPGIS instruments that was previously lacking. The Park Life PPGIS elicits information on park use that is essential for public health and physical activity researchers, urban planners, and park managers to develop informed planning and public health policies and programs that promote the use of parks.

CONFLICT OF INTEREST STATEMENT

The authors declare no competing interests.

ETHICS STATEMENT

The University of Western Australia’s Human Ethics Committee provided ethics approval (2019/RA/4/1/8734).