Associations between the built environment and jogging behaviour based on voluntarily contributed tracking data: a systematic review and meta-analysis

Abstract

Background

Running is a widely promoted form of physical activity with significant public health benefits, yet the built environment influences its engagement. Existing evidence on the associations between the built environment and running behaviour remains heterogeneous, with prior reviews not quantifying the overall effect sizes. Additionally, gaps persist in understanding how emerging geospatial data, such as volunteered geographic information (VGI), can enhance insights into runnability.

Aim

To provide (1) a comprehensive synthesis of the literature and meta-analysis of the evidence for the effects of the built environment on jogging behaviour, and (2) to identify methodological limitations and future research priorities for promoting running-inclusive cities.

Methods

Three databases (the Web of Science Core Collection (WoS), Scopus, and PubMed) were systematically searched for English-language studies published up to December 31, 2024. Meta-analysis was conducted to obtain pooled elasticity values for the environmental factors.

Results

Of the 1,884 studies identified, 39 studies fulfilled the inclusion criteria, and 14 studies were suitable for meta-analysis leveraging VGI-derived physical activity data. Meta-analysis revealed that floor area ratio had the largest effect size, followed by land use mix and blue space density. Distance to parks and public transport density showed minor effects. Natural environment features (e.g., blue space density and green view index) consistently correlated positively with running activity, while terrain slope exhibited context-dependent relationships. Critical methodological limitations included insufficient spatiotemporal analysis, overreliance on single-platform VGI data, and inconsistent geographic units.

Conclusions

To advance runnability research, future studies should adopt dynamic spatiotemporal modelling, integrate multi-platform VGI with participatory GIS, and employ equity-focused metrics and demographic-stratified analyses. These strategies will inform evidence-based urban planning to create running-inclusive environments, ultimately supporting population health through targeted built environment interventions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12942-025-00428-4.

Introduction

Running, health, and the environment

Promoting physical activity is crucial for preventing non-communicable chronic diseases, reducing all-cause mortality, and contributing to the economic and environmental sustainability of cities [1–3]. Running, as a non-organised and cost-effective form of physical activity, has been associated with a range of physical and mental health benefits, including reduced resting heart rate, improved lung function, increased longevity, and a lower risk of depression and anxiety [4, 5]. Beyond these well-documented health benefits, running has also emerged as a zero-carbon and sustainable commuting mode [6, 7]. Given its health and environmental benefits, running has attracted increasing research attention, with studies examining its spatial and temporal patterns and identifying environmental factors that promote this activity [8–11]. While the built environment’s relationship with walking [12] and cycling [13] is well-established, running warrants a distinct assessment [14–19]. As a vigorous-intensity activity, running offers unique, efficient health benefits for combating non-communicable diseases [4]. Furthermore, its spatial patterns and environmental demands differ significantly [20]; runners typically traverse longer distances than walking and may prioritize extended, uninterrupted routes over the fine-grained connectivity important for walking [12] or the dedicated infrastructure essential for cycling [13]. A synthesis specific to running is therefore critical to inform targeted urban design strategies that support this distinct form of physical activity. Understanding these relationships is essential for informing urban planning and health policy aimed at creating activity-supportive environments.

Defining runnability with user-generated geospatial information

Previous research on the association between the built environment (e.g., street connectivity, proximity to green spaces, land use mix, and urban density) and running has primarily relied on surveys rather than tracking data techniques [20]. While surveys can effectively capture general patterns, such as running frequency, perceived attractiveness of the running environment, and restorativeness [16, 17], they lack the spatial precision needed to correlate running activities with specific built environment characteristics. With the advancement of digital participatory mapping approaches, Public Participation Geographic Information Systems (PPGIS) have gained traction to collect non-expert spatial data on running routes retrospectively reported by participants [21]. For example, D Huang, B Jiang and L Yuan [20] found that nature exposure was associated with perceived satisfaction with running routes in the Helsinki metropolitan area, and that the built environment-running associations likely vary spatially [22]. Although PPGIS provides an in-depth understanding of running geographies, its data collection faces challenges such as representativeness issues, time-intensive processes, and the routes being accurately recalled [23].

To address some of these limitations, location-acquisition technologies, including global positioning system (GPS)-enabled devices (e.g., wearables and smartphones), have become widely adopted, contributing to the growth of volunteered geographic information (VGI) [24]. VGI refers to citizens’ voluntary contribution of geographic data [24, 25]. Key strengths of VGI data are cost-effectiveness, richness, and the capacity to provide frequent updates. However, concerns about data reliability, uneven spatial completeness, and privacy issues remain notable limitations [26]. User-generated sport-tracking platforms (e.g., Strava, Keep) offer large-scale, real-time data with high spatial and temporal resolution on running routes. Numerous studies have examined the relationship between the built environment and running activity using crowdsourced GPS data from Strava [8, 27], Keep [28, 29], Edooon [9], and Endomondo [30], yielding partially contradictory results with various effect sizes. Therefore, identifying the environmental correlates most consistently associated with increased running activity is essential for guiding and prioritising urban planning.

Related work

A few reviews have summarised the associations between the built environment and physical activity (e.g., walking, cycling, and running). For example, PRW McCrorie, C Fenton and A Ellaway [31] used a combination of GPS, GIS, and accelerometry to highlight the relationship between the environment and physical activity in children and adolescents. However, the reviewed studies primarily focus on total physical activity, rather than isolating specific forms of physical activity such as running. Y Yang, X Wu, P Zhou, Z Gou and Y Lu [13] reviewed survey-based studies published between 2007 and 2017 that examined associations between cycling behaviour (e.g., frequency and duration) and the built environment at participants’ residential locations. However, this review does not account for cycling activity beyond the residential neighbourhood, creating a critical gap in our understanding of how specific environmental characteristics influence physical activity patterns at the precise locations where they occur. Our review addresses this limitation by focusing on VGI-based studies that provide geospatial data on runners’ environmental preferences, revealing how specific urban features (e.g., park trail density, street connectivity) influence route selection behaviour at fine spatial scales. A scoping review of 102 studies examined environmental influences on runnability, categorising relevant factors into six domains: nature exposure, perceived safety, traffic conditions, pollution, terrain, and connectivity [32]. While this comprehensive review analysed both individual running behaviours (e.g., frequency, motivation, perceived restorativeness) and spatial activity patterns at specific locations (e.g., street segments or parks), it did not conduct a meta-analysis to pool the effect sizes of different built environment factors across studies.

While previous reviews have offered qualitative summaries of the built environment’s relationship with jogging [32], a quantitative synthesis capable of guiding evidence-based policy has been lacking. This study presents, to our knowledge, the first systematic review and meta-analysis to fill this gap. Our primary contribution is the calculation of pooled effect sizes (elasticities), which transcend mere statistical significance to serve as actionable tools for urban planning and public health. These elasticities enable policymakers to move beyond generic recommendations by identifying which specific environmental features—such as greenspace exposure, street connectivity, or population density—provide the strongest leverage on jogging behaviour, thereby facilitating cost-effective investment in health-promoting environments. However, the evidence base relies heavily on VGI platforms like Strava, which have a well-documented demographic bias toward younger, male, and more athletic populations [33], creating a critical equity blind spot. Therefore, the aims of this review are (1) to synthesise the existing evidence narratively and quantitatively through a meta-analysis and (2) to evaluate limitations, including demographic representativeness, to identify priorities for future research.

Methods

Search strategy

The review followed the PRISMA guideline [34]. Our literature review used three major databases: the Web of Science Core Collection (WoS), Scopus, and PubMed. We extracted studies from the databases on January 26, 2025. We systematically searched for articles using the following combination of search terms applied to the title, abstract, and keywords: (“run” OR “running” OR “runnability” OR “jogging”) AND (“environment” OR “greenspace” OR “green space” OR “nature”) AND (“association” OR “relationship” OR “correlation” OR “associated” OR “related” OR “correlated”) AND (“GIS” OR “VGI” OR “spatial” OR “geographical”). The search was restricted to peer-reviewed original research articles published in English between 2010 and 2024 (Table 4). This date restriction was applied because VGI-based physical activity studies only became feasible following the advent of GPS-enabled fitness apps like Strava (founded in 2009) and Keep (2014), which enabled crowd-sourced activity data collection.

Table 4.

Search strings and their setup in WOS, scopus and pubmed used in this review

Basic search string	Number of studies			In string duplicate removed	Exact string WOS	Exact string scopus	Exact string PubMed
	WOS	Scopus	PubMed
(run OR running OR runnability OR jogging) AND (environment OR greenspace OR green space OR nature) AND (association OR relationship OR correlation OR associated OR related OR correlated) AND (GIS OR VGI OR spatial OR geographical)	872	873	139	716	((((((((TS = (run OR running OR runnability OR jogging) AND TS = (environment OR greenspace OR green space OR nature) AND TS = (association OR relationship OR correlation OR related OR associated OR correlated) AND TS = (GIS OR VGI OR geographical OR spatial))))))) AND LA=(English)) AND DT=(Article)) AND PY=(2010–2024)	TITLE-ABS-KEY(“RUN” OR “RUNNING” OR “RUNNABILITY” OR “JOGGING”) AND TITLE-ABS-KEY(“ENVIRONMENT” OR “GREENSPACE” OR “NATURE” OR “GREEN SPACE”) AND TITLE-ABS-KEY(“ASSOCIATION” OR “RELATIONSHIP” OR “CORRELATION” OR “related” OR “associated” OR “correlated”) AND TITLE-ABS-KEY(“GIS” OR “VGI” OR “geographical” OR “spatial”) AND PUBYEAR > 2010 AND (LIMIT-TO (DOCTYPE,“ar”)) AND (LIMIT-TO (LANGUAGE,“English”))	((((((RUN[Title/Abstract] OR RUNNING[Title/Abstract] OR RUNNABILITY[Title/Abstract] OR JOGGING[Title/Abstract]) AND (ENVIRONMENT[Title/Abstract] OR GREENSPACE[Title/Abstract] OR NATURE[Title/Abstract] OR (“GREEN“[Title/Abstract] AND “SPACE“[Title/Abstract])) AND (ASSOCIATION[Title/Abstract] OR RELATIONSHIP[Title/Abstract] OR CORRELATION[Title/Abstract] OR ASSOCIATED[Title/Abstract] OR RELATED[Title/Abstract] OR CORRELATED[Title/Abstract])))) AND (GIS[Title/Abstract] OR VGI[Title/Abstract] OR GEOGRAPHICAL[Title/Abstract] OR SPATIAL[Title/Abstract])) AND (“2010/01/01“[PDAT] : “2024/12/31“[PDAT])) AND (Journal Article[ptyp])

After removing 716 duplicates, 1,168 unique articles were identified. The article selection process involved two stages of screening (Fig. 1). First, we screened the abstracts to exclude articles that met any of the following criteria: (1) the study focused on irrelevant topics, or (2) the study did not examine the associations between running activity and any built environmental characteristic. Built environmental characteristics were categorized according to established frameworks [35] focusing on natural environment (e.g., green space density, NDVI, green view index); urban design and transportation (e.g., intersection density, street density, public transport accessibility); and land use (e.g., land use mix, population density, point-of-interest density). Variables related to perception (e.g., safety, maintenance) were analysed separately due to methodological differences in their measurement. Other environmental variables, such as weather conditions, were excluded as they fall outside the scope of this review, which focuses specifically on the built environment. This resulted in excluding 1,137 articles, leaving 31 articles for full-text review. Second, we performed a backwards and forward citation search (i.e., snowballing) based on the reference lists of the eligible articles. This resulted in an additional 22 relevant articles. Fifty-three articles were subjected to full-text evaluation, during which articles were excluded if they: (1) did not specifically measure running as a distinct form of physical activity but instead aggregated it with other activities such as walking or cycling (n = 10), (2) did not use crowdsourced GPS data (e.g., from fitness trackers, smartphones, or navigation apps) to quantify running activity (n = 4). After applying these exclusion criteria, 39 articles were deemed eligible for further analysis. The lead author performed the initial title and abstract screening, with verification by the second and fourth authors. For full-text evaluation, the lead and second author independently assessed eligible studies, extracted relevant data, and synthesized findings under the oversight of the fourth author. Any disagreement was resolved through discussion until a consensus was reached.

Fig. 1.

The selection process for the literature review, detailing the number of studies included and excluded at each stage

Data extraction and synthesis

Following previous reviews on walking [12] and cycling [13], we extracted key information from each study, including author details, year of publication, location of the study area, sample size, analysed geographic unit, buffer size, types of environmental attributes, statistical models, running activity-related outcome variables, the reported associations, number of crowdsourced data records, the source GPS data, and the journal. The extracted data were tabulated in Table 5 (for detailed associations, see Supplementary Table S1 due to space constraints) and analysed using frequency analyses and cross-tabulation in Excel and SPSS (version 26).

Table 5.

Characterization of studies by country, sample size, crowdsourced data records, built environmental factors, analysed geographic unit, outcomes, and major results

Author		Country	Sample size	Crowdsourced data records	VGI data source	Built environment factors	Analysed geographic units	Outcomes	Results
1	Y Chen, B Wang, J Huang, H Gao and X Shu [18]	China	NA	5767 trips	Foooooot	1. Density of services 2. Diversity 3. Accessibility 4. Terrain 5. Landscape	Hexagonal grid	Running distance	Land use diversity and varied topography are appealing for both leisure trips and trip length
2	L Dong, H Jiang, W Li, B Qiu, H Wang and W Qiu [14]	United States of America	26,201	Heatmap	Strava	Macro-scale built environment (BE) Density: building density, 311 calls density, reported safety concerns density Diversity: POI density, land use Design: tree density, open space area, street segment length, street light density, street slope, traffic signal density Destination accessibility Distance to transit Eye-level street environment (SE) Objective features Building, sky, tree, road, sidewalk, person, car, fence, signboard, street light, traffic light, wall, grass, plant, railing, earth, awning, ashcan, ceiling, water, column, flower, mountain, van, bridge, door, boat, windowpane, chair, minibike, bicycle, fountain, field, lake, sculpture, lamp, pier, skyscraper, hovel, bulletin board, animal, booth, sofa, glass, desk	Sample points	Running intensity	The street environment is significantly correlated with running. Accounting for the spatial effects, the collective strength of street attributes was almost the same as the counterpart of the built environment Street factors can complement built environment factors, indicating the necessity of using both macro-scale and eye-level environmental features to interpret running In addition to higher accessibility and more public transportation, the safer, wider and relatively open streets with more natural views, street lights, amenities and furniture, could promote running, while the enclosed environment, dense and overwhelming buildings, excessive interruptions on streets might hinder running
3	C Fang, R Homma and T Qiu [43]	Japan	69,918	46,628 points	Strava	Population density Residential density Land-use mix River area Tree canopy area Elevation Road slope Road density Distance to the city center Distance to the local hub Distance to the living hub Distance to school playgrounds Distance to natural parks Located in large-scale park Located in small-scale park Distance to river Distance to railway stations Distance to bus stops	Road segments	Running intensity	Large parks were the primary places for recreational running Population density and small-scale parks significantly affected leisure running on weekdays Road density and distance to rivers significantly affected weekend leisure running
4	T Fang, L Zhou, Z Cai, Z Tan, C Chen, J Zheng and C Fang [11]	China	87	22,001 data records	Keep	Total area NDVI Blue landscape LSI Road length Road density Road width Synthetic surface track Number of public toilets Density of intersections Road density Residential land ratio Population density Distance to the city center Average night light Density of bus stops Building land ratio Commercial land ratio	Parks	1. Running intensity 2. Running distance 3. Running duration 4. Running speed	Total area and road length positively correlated with the running flow, while distance to the city center and NDVI negatively correlated with the running flow Running distance shows positive correlations with the number of public toilets and road length, while it is negatively correlated with density of bus stops nearby and road density nearby Running time is positively correlated with number of public toilets and road length, and negatively correlated with density of intersections nearby and density of bus stops nearby
5	F Gao, X Chen, S Liao, W Chen, L Feng, J Wu, Q Zhou, Y Zheng, G Li and S Li [77]	China	NA	56,471 jogging tracks	Keep	Road accessibility Public transport Jogging track Facility density Facility diversity Sports facility density Park Street lighting Green View Index Sky view factor Safety score Wealth score Temperature Slope Air quality	Road segments	Running intensity	Road accessibility, and public transport density were found to be significantly and positively correlated with jogging frequency Sports facilities were found to be positively correlated with jogging Outdoor street night lighting was positively correlated with jogging GVI and perceived safety were found to be positively correlated with jogging
6	X Gu, Z Lai, L Zhu and X Liu [28]	China	12,096	NA	Keep	Green View Index Normalized Difference Vegetation Index Cooling Effect Population Density Aging Degree Building Density Building Spacing Road Density Road Facilities Land-use Mix Sky View Index Visual Motorization Index Terrain Slope Air Humidity	Grids	Running intensity	Areas with higher cooling effects and street-scale greenery correlate with increased jogging vitality
7	H Guo, S Zhang, Y Liu, R Lin and J Liu [47]	China	882	545 routes with a total of 9.73 million running times	Keep	Macroscale built environment Population density Average building height Land use HHI Road connectivity Density of road intersections Distance to nearest metro Density of bus stops Distance to nearest park Distance to nearest river Microscale streetscapes Green view index Crowd concentration index Sky view factor Building-to-street ratio Public-facility convenience index Non-motorized vehicles interference Street complexity Visual crowdedness Transparency	Road segments	Running intensity	1) Sky and green view indexes were positively associated with running intensity, whereas visual crowdedness had a negative effect; 2) There were negative interactions of land use Herfindahl–Hirschman index with sky and green view indexes, while a positive interaction was observed for visual crowdedness
8	W Guo, J He and W Yang [64]	China	NA	287,663 trajectory trips with 96,939,911 GPS points	Edooon	PM2.5 exposure Spatial factors (six administrative districts) Meteorological factors (Temperature, Precipitation, Wind Speed)	Grids	1. Activity space size 2. Running distance 3. Running duration 4. Running rotation 5. Running eccentricity 6. Running speed	(1) There exist significant spatiotemporal disparities in jogging exposure to PM2.5. Joggings in the city center, in the morning, on weekdays and in autumn and winter seasons were exposed to higher pollution concentrations (2) Jogging behavior characteristics, especially distance, activity space size, duration and rotation, were systematically associated with PM2.5 exposure across space and time (3) The role of gender directly shaped joggers’ dose inhalation of PM2.5 pollution and indirectly via duration, timing choice and distance (4) The effects of weather conditions on joggers’ exposure to PM2.5 are mainly via direct effects, whereas the direct impacts of precipitation and wind speed are mitigated by indirect effects stemming from jogging behavior patterns
9	SR Harden, N Schuurman, P Keller and SA Lear [55]	Canada	242,265	NA	Strava	Green and/or Blue Space SES Urbanicity	Road segments	Running intensity	High neighborhood SES, the presence of green and/or blue space, and high population density are associated with increased running activities in all age and gender groups
10	D Huang, F He and W Liu [8]	China	9,392	NA	Strava	Connectivity Terrain slope Blue space density Distance to the city centre Levels of PM2.5 concentrations Normalized difference vegetation index Trail density Landscape shape index Green view index Population density Building density Street density Public transportation node density Housing price	Park trails	Running intensity	Trail network connectivity was the only environmental attribute indicating consistent and positive associations with running intensity Blue space density was positively correlated with running intensity in urban parks but indicated no significant association in forest parks before the pandemic In the pre-pandemic era, population density was positively related to running intensity in urban and forest parks. However, after the pandemic, the associations between running behaviours and population density remained positive in forest parks but turned insignificant in urban parks. The outbreak of the pandemic also altered the influence of other park features (e.g. park shape and trail density) on running intensity
11	D Huang, M Tian and L Yuan [35]	Finland	13,322	NA	Strava	Green View Index NDVI Blue space density Terrain slope Floor area ratio Height-to-width ratio Land use mix Traffic noise Air quality index Street density Traffic volume Traffic accident density Street network connectivity Population density Median annual household disposable income	Road segments	Running intensity	Street greenery assessed by GVI represented the greenness exposure to runners better than top-down greenness assessed by NDVI, and thus can be considered as a more reliable predictor for running behaviours Blue space density was the predominant factor and associated with running intensity positively Running intensity negatively correlated with urban density, connectivity and traffic accidents, and positively correlated with traffic noise and air pollution Population density and income level were positively associated with running intensity
12	H Jiang, L Dong and B Qiu [41]	United Kingdom	40,290	NA	Strava	Macro-scale built environments 5Ds Population density Job density Building density Street type POI entropy Open space area Canopy density Number of intersections Number of traffic lights Number of parking lots Maximum speed Street length Number of crimes Number of traffic accidents Number of fires Annual mean NO2 Annual mean PM2.5 Street noise pollution level Street slope Night-light intensity Annual mean temperature Destination accessibility Distance to transit Micro-scale built environments Wall, building, tree, road, grass, sidewalk, earth, plant, car, fence, signboard, awning, streetlight, van, ashcan, railing, person, minibike, chair, sculpture, bicycle, column, bridge, water, fountain, windowpane, mountain, ceiling, booth, sofa, lamp, skyscraper, lake, bulletin board, desk, pier Sky view factor (SVF)	Sample points	Running intensity	(1) Running occurs more frequently on trunk, primary, secondary, and tertiary roads, cycleways, and footways, but runners choose tracks, paths, pedestrian streets, and service streets relatively less; (2) Safety, larger open space areas, and longer street lengths promote running; (3) Streets with higher accessibility might attract runners; (4) Higher job density, POI entropy, canopy density, and high levels of PM 2.5 might impede running (5) Wider roads, more streetlights, trees, higher sky openness, and proximity to mountains and water facilitate running; (6) More architectural interfaces, fences, and plants with low branchingpoints might hinder running
13	Y Lai, J Li, J Zhang, L Yan and Y Liu [71]	China	8,979	660,891 trajectories	Keep	Population density Inbound distance Outbound distance Local residents POI density POI diversity POI completeness Education facility Cultural facility Health facility Sports facility Grocery facility Street quality Green space	Grids	Running intensity	The results reveal thestrong impact of urban form on local vibrancy but not physical activeness Although physical activity is related to the local population, residential percentage, business, and public facilities, it does not show a strong spatial gradient that matches the monocentric urban form as vibrancy
14	B Li, Q Liu, T Wang, H He, Y Peng and T Feng [46]	China	735	NA	Smartwatch providers (Duorui wearable and application)	Land-use mix Residential land density Building density Arterial road density Secondary road density Branch road density Bus stop density University campuses density Comprehensive parks density Community parks density Square density Market density Living service points density NDVI	Sample points	Running intensity	Built environment impacts on outdoor physical activities in Changsha are not always consistent with similar studies’ results in other cities.The most effective measures to promote outdoor physical activities are the provision of good arterial and secondary road networks, and community parks
15	Y Liu and Y Lai [67]	China	NA	11,080 trajectories	Keep	Streets Tracks	University campus	Spatial-temporal patterns	Based on classification and pattern identification, the results reveal three major running activity types on streets, tracks, and mixed spatial conditions
16	Y Liu, J Hu, W Yang and C Luo [39]	China	147	33,738 trajectories	Edooon	Park size NDVI LSI Presence of walking loops Presence of waterscape Park trail density Distance to CBD Number of road intersections Number of bus stops Number of population Number of SFPOIs	Parks	Running intensity	Walking loops and waterscapes have positive effects on jogging flow The landscape shape index of urban parks and the distance to the city center negatively affect the jogging flow
17	Y Liu, Y Li, W Yang and J Hu [56]	China	2,903	96,850 trajectories	Edooon	NDVI GVI Openness Number of parks Number of tracks Distance to the nearestpark Distance to the nearesttrack Distance to the nearest water body Number of road intersections Number of bus stops Distance to the nearest bus stop Population density Building density	Blocks	Running intensity	(1) BE factors exert diverse nonlinear effects on jogging at trip and OD levels (2) Quantity and accessibility of facilities contribute largely to model predictive power (3) Nonlinear effects are symmetrical for O/D of jogging, unlike long-distance travel. Distance to park, distance to track, and population density show U-shaped effects on OD volume (4) Effective ranges and thresholds in nonlinear effects vary across trip/OD levels
18	Y Liu, G Zhang and W Yang [78]	China	NA	7,269 trajectories	Edooon	Temperature Wind speed Relative humidity Direct solar radiation Solar scatteredradiation Total radiation Building footprints and heights Elevation Land use Road network	Traffic analysis zones (TAZs)	Spatial-temporal patterns	Jogging patterns influence the heat exposure. Regarding motion patterns, looping jogging has lower sum heat exposure than non-looping. From temporal dimensions, weekend joggers face higher heat exposure than weekdays. For spatial patterns, joggers along east-west experience higher heat exposure than along a south-north direction; Roadside jogging exhibits the lowest sum heat exposure and park jogging demonstrates the highest one because of the duration of jogging Heat emission, building shade, and physical activity collectively affect heat exposure
19	C Luo, H Yu, Y Liu and W Yang [45]	China	2,420	83,332 trajectories	Edooon	UGS area Per capita UGS area NDVI UGS trail density Water area ratio Average slope Intersection density Number of bus stops	Blocks	1. Running intensity 2. Running duration	The findings revealed distinct spatial imbalances in the supply and demand of PA services in the case of Chongqing PA services exhibited spatial heterogeneity, with undersupply in the urban center and oversupply inthe suburbs
20	P Luo, B Yu, P Li and P Liang [79]	China	NA	NA	Strava	FAR Land-use mix River line length Road density Green space area Number of bus stations Number of enterprises Green view index Sky view index Road view index	Grids	Running intensity	GVI has a wide-ranging and positive impact on running activities while hampers them in the PA-unfriendly old town The road view index widely promotes running activities but hinders them in some areas of the old town dominated by automobiles and under construction
21	P Norman and CM Pickering [73]	Australia	40	96,720 running routes	Strava, MapMyFitness and Wikiloc	Park size Average slope Highest elevation Elevation range Trail length Recreational trail length Mountain biking trail length Walking and running trail length Rainforest area Total rainforest area Direct distance to large urban area boundary Direct distance to large urban area centre Road distance from park facilities to urban areas	Parks	Running intensity	Distance (road and directdistance) from large urban areas best predicted visitation using routes from MapMyFitness and Strava with three parks within 2 km of urban areas experiencing over 70,000 visits (Strava) For urban parks, recreational trail length best predicted usage, while for more remote parks, the direct distance to urban areas remained the most important factor. In contrast, people using the adventure platform Wikiloc preferred more remote parks with rugged terrain The results highlight factors affecting park popularity including distance
22	P Norman, CM Pickering and G Castley [10]	Australia	226	226 running routes	MapMyFitness/on-ground trail counters	Their relative proximity to urbanareas Ease of access The types of activities permitted The availability of facilities	Park trails	Running intensity	Mountain biking was more popular than walking or running across the three reserves Bikers went further, used a greater range of trail combinations and used more reserves per trip than the other two activities
23	T Shi and F Gao [80]	China	3,738	56,471 trajectories	Keep	Population density Road density Public transport Facility density Facility diversity Sports facility density Residential density Distance to park Nighttime lighting Building density Green view index Sky view index Wall Perceived safety Perceived liveliness Temperature Air quality Slope Distance to body of water	Grids	Running intensity	(1) Factors related to the built environment and street perception significantly influence jogging frequency distribution (2) Public sports facilities, the level of greenery, and safety perception were identified as key factors influencing jogging activities, representing the three aspects of service facilities, objective perception, and subjective perception, respectively (3) Specifically, the influence of each factor on jogging activities displayed significant spatial variation. For instance, sports facilities and greenery level were positively correlated with jogging frequency in the city center
24	L Tan, J Jiang, M Guo and Y Zhong [29]	China	1,063	NA	Keep	Environmental perception factors Safety Vibrancy Cleanliness Slope Width Accessibility Connectivity Air quality Scenic beauty Greenery Lighting Soundscape Temperature Environmental elements factors Vehicles Pedestrians Traffic infrastructure Landscape Green space Waterfront space Service facilities Lighting facilities Pavementculture Buildings Vertical elements	Land use type	Running intensity	Streets, residential areas, campuses, parks, and greenways possess significant potential to support jogging, particularly streets These types of land use exhibit varying spatial potentials and attractions in environmental preferences
25	N Wang, W Wei, Y Qian, H Gao and H Qiu [63]	China	35,662	220,188 pieces of data	Strava	Park view elements Wall, building, sky, floor-road-earth, tree, grass-plant, sidewalk-path, person, car, water, fence-railing, signboard, bench, streetlight-pole, ashcan, altitude and slope gradient	Sample points	1. Running HR 2. Running speed	(1) The influence of the same park view elements on the exercise physiological indicators of different genders is small; (2) Park view elements combination based on sky, grass-plant and tree can better stabilize the walking HR of the older adult; (3) Semi-enclosed trail dominated by tree can improve the walking HR and speed of people with larger body weight; (4) Natural routes dominated by sidewalk-path and supplemented by tree and sky elements are more suitable for walking, while the trails with larger sky area, no trees and wider trails are more suitable for running
26	M Xie, Z Feng, W Long, S Wang, X Liu, G Ji and X Guo [52]	China	10,664	3,271 trajectories	Keep	NDVI Terrain slope Blue space distance GVI SOI Road intersection density Traffic signal density Public transportation station density Street density POI density Building density Building height Housing price Population density	Road segments	Running intensity	Streets with higher running intensity are closer to water bodies and green spaces, while lower-intensity areas are often associated with busy traffic, noisy commercial activities, and sparsely populated regions
27	L Yang, B Yu, P Liang, X Tang and J Li [81]	China	2,788	NA	Strava	Residential building density FAR Land-use mix Water area River line length Green space area Light index Road density Number of bus and metro stations Number of enterprises Number of schools Number of stadiums Housing price	Grids	Running intensity	Land-use mix has insignificant effects on the physical activity of residents Road density, water area, green space area, number of stadiums, and numberof enterprises significantly facilitate running River line length and the light index have positive associations with running Housing price is positively correlated with running
28	W Yang, H Chen, J Li, W Guo, J Fei, Y Li and J He [48]	China	1,085	454,575 trips	Edooon	Greening view index Visual openness Visual enclosure Visual walkability Visual humanization index Visual entropy index Population density Building density Road density Intersection density Land-use mix NDVI Distance to bus stop Distance to metro stop Distance to track Distance to park Distance to waterway Number of tracks Number of parks Number of waterways	Traffic analysis zones (TAZs)	Running intensity	Visual greenery, openness, and walkability play an important role in determining jogging behaviors Effects of some VE variables display distinct scale effects and spatial non-stationarity Notably, the dual effects or synergistic effects of specific VE factors, such as visual greenery, walkability, between inside and outside the TAZ were identified Effects of visual greenness and openness, on jogging behavior vary across peak-hours, weekdays and weekends, and seasons, respectively
29	W Yang, J Fei, Y Li, H Chen and Y Liu [9]	China	1,851	648,010 trips	Edooon	Macroscale BE variables Population density Building density Land-use mix Intersection densityroad density The distance to the nearest track, park, and waterway The number of tracks, parks, and waterways distance to the nearest bus stop and metro stop NDVI Greening view index Sky view index Visual motorization index visual humanization index Shannon’s diversity index	Traffic analysis zones (TAZs)	Running intensity	(1) Macroscale BE including sports amenities and accessibility play an important role in affecting jogging (2) All BE factors have nonlinear association with jogging, and the effective range and threshold effects vary by variables. Notably, several variables associated with jogging are in inverted U or V shapes (3) The complex interaction effects including synergistic, weakened, among BE factors were scrutinized. BE factors regarding accessibility and sports amenity interact more easily
30	W Yang, J Hu, Y Liu and W Guo [49]	China	5,942	80,507 trajectories	Edooon	GVI SVF Number of sport and leisure POIs Number of retail POIs Number of bus stops Number of road intersections Walking time to the nearest park Proximity of road to waterscapes Location at intervals of ring roads Population heat index FAR	Road segments	Running intensity	(1) BE factors, including sky view factor, bus stop density, presence of waterscapes, and geographic location significantly, impact jogging activity (2) The significance and effect of BE factors vary across time. Jogging activities are more sensitive to BE on weekends than on weekdays (3) Jogging trips are more closely related to BE factors in urban areas than in the suburbs. Jogging activities are mainly affected by the artificial environment in urban areas and the natural environment in suburban areas. Moreover, the effects of multiple BE factors become obvious as trip distance increases
31	W Yang, Y Li, Y Liu, P Fan and W Yue [50]	China	1,851	575,070 trips	Edooon	Population density Building density Land-use mix Intersection density Road density The distance to the nearest track, park, and waterway The number of tracks, parks, and waterways distance to the nearest bus stop and metro stop Greening view index Sky view index Visual motorization index visual humanization index Simpson’s diversity index Housing price Rent price	Traffic analysis zones (TAZs)	Running intensity	(1) Built Environment (BE) factors play a more important role than visual landscape factors in determining jogging behavior, with sports facilities such as tracks and parks having higher contributions (2) All environmental variables, including the BE, visual landscape, and social economics, exhibit nonlinear and threshold effects on jogging behavior (3) Certain factors such as population density, number of parks, Greening View Index, and Sky View Index exhibit differential effects across different periods, regions, and mobility patterns of jogging
32	C Zhang, D Shi and Z Xiao [53]	China	3,568	1,371 trajectories	2bulu	Elevation Slope NDVI Green space area Water body area Population density FAR Retail & Service Facilities Density Land Use Mix Integration Index Choice Index Intersection Density Living Road Density Traffic Road Density	Grids	Running intensity	Natural environmental factors are most contributing to outdoor jogging, while density-related built environment factors contribute the least. Additionally, environmental effects vary in scale, direction, and intensity, with seven variables exerting global impacts and five showing localized effects
33	H Zhang, S Nijhuis, C Newton and Y Tao [68]	Netherlands	24,050	35,685,390 records	Strava	Types of blue space Water View Index GVI Openness Building density Land use mix Visual complexity Road traffic Integration Choice Urbanicity Non-western Low-income	Road segments	Running intensity	Recreational exercise levels on street segments vary based on the blue space type and design. Compared to inland canals and rivers, small-scale recreational waterbodies are more conducive to running but not cycling Interestingly, the Water View Index shows a general negative association with both activities after adjusting for the blue space type. Higher GVI, lower building density, more diverse land use, greater connected street network and fewer traffic elements, are associated with more running and cycling exercises
34	S Zhang, Z Wang, M Helbich and D Ettema [30]	Netherlands	Urban: 25,720 tracks, Rural: 8,140 tracks	66,497 tracks	Endomondo	The level of greenness Residential building density Land use mix Street connectivity Urbanisation index The proportion of water bodies	Residential (1,000 m) and GPS-based buffer (25 m)	1. Spatial-temporal patterns 2. Running duration 3. Distance 4. Frequency	Both urban and rural runners preferred environments with more blue spaces and fewer streets, and a lower density of residential buildings More green and blue spaces, streets, less addresses and land use with higher heterogeneity were in the surroundings of > 30 min tracks
35	S Zhang, N Liu, B Ma and S Yan [42]	China	959	0.973 million check-ins	Keep	Resident number Point-of-interest Green-land and river Road network Resident density Function density Function mix Green-land distance Blue space distance Waterfront trail section TSV sky TSV building TSV tree Old façade Road section length Road section width Junction density Intergration Wall continuity Building density	Road segments	Running intensity	Blue space and trail continuity are the most important factors in improving road running intensity There is an optimum design value for the sky openness and the street enclosure, which need to be balanced with shade while meeting the light of the road. And it is also important to provide appropriate visual permeability. Furthermore, unlike daily activities, it was found that higher function mixture and function density did not have significant positive effects on the road running intensity
36	Q Zhong, B Li and Y Chen [54]	China	8,311	NA	Dorray	Residential land density Green land density Nighttime light Arterial road density Secondary road density Secondary road density Land-use mix Facilities diversity Metro line density Bus stop density Precipitation Temperature NDVI Slope Bus distance density Metro distancendensity Population density Gross Domestic Product	Route buffer	Running intensity	(1) jogging activities (JA) were mainly concentrated in the community footpath environment; (2) Built environments, social environments, and natural environments significantly affected the relative jogging distance of residents; (3) Arterial road density (ARD) and bus distance density (BDD) have opposite significant effects on the JA of communities and green land footpaths; (4) ARD has the significant opposite effect on the JA for residents of different genders on urban footpaths and community footpaths. Facilities diversity (FD), population density (PD), and bus stop density (BSD) also had significant opposite effects on the JA of residents of different genders on greenland footpaths
37	Q Zhong, B Li and T Dong [40]	China	NA	4,483 trajectories	Sport app (not specify)	Population density Building density Water density Residential land density Green space density Land use mix Urban functions mix Road density Road crossings density Road connectivity Nighttime Lighting Index Degree of relief NDVI Distance to bus stations Bus line density Bus stop density Accessibility to recreational facilities, catering facilities, commercial facilities, workplace, educational facilities, government organizations and scenic spots	Grids	Running intensity	Women participated in walking, jogging, and bicycling activities at a higher rate than men Various leisure-time physical activities have different requirements for the built environment
38	Y Zhong, M Guo, M Zhang and L Tan [19]	China	1,001	NA	Keep	Pedestrians Green space Waterfront space Landscape Sports facilities Commercial area Accessibility	Route buffer	Running intensity	(1) Discrepancies between expert opinions and public attention, with experts more likely to overlook factors such as slope, visual quality, width, and acoustic quality; (2) The public tends to notice factors like pedestrians, green spaces, waterfront spaces, landscapes, culture, sports facilities, and accessibility on routes with high jogging frequency; (3) EFPAI for traffic infrastructure, waterfront spaces, landscapes, residential areas, and campuses is associated with route shape;
39	W Zhou, Z Liang, Z Fan and Z Li [44]	China	31,356	201,785 trajectories	Keep	Population density Building density Street density Road intersections Functional density POI entropy Street types Bus stops Subway stations Park green space Greenways NDVI	Road segments	Running intensity	Urban park green space, greenway, and the normalized difference vegetation index had the most significant effects on running activity The effects of population, buildings, streets, road intersections, and points of interest on running activity changed during the Coronavirus disease 2019 pandemic

Meta-analysis using weighted average elasticity values of built environmental characteristics

We computed elasticity values for the environmental factors in each study to assess how sensitive the dependent variable is to changes in the relevant independent variables [13, 36, 37]. For example, if the elasticity of street connectivity is 0.4, it indicates that a 1% increase in street connectivity is associated with a 0.4% increase in running activity.

As done by R Ewing and R Cervero [38], elasticity values were derived using one of two approaches: (1) directly extracting reported elasticity values from the reviewed studies, or (2) calculating elasticity using regression coefficients along with the mean values of the dependent and independent variables. The formulas for these calculations are provided in Table 1, depending on the type of regression model applied to estimate the coefficients. We calculated weighted average elasticity values for each environmental characteristic using sample size as the weighting factor, but only when the following three conditions were met: (1) at least three studies reported relevant data; (2) descriptive statistics were available for both the dependent and independent variables; and (3) a statistically significant association was found between running activity and the environmental characteristic. Strong evidence was defined as cases where the number of statistically significant expected results was more than twice the combined total of unexpected and null findings. Evidence was considered emerging when expected results outnumbered the combined count of unexpected and null findings but did not meet the threshold for strong evidence [13].

Table 1.

Approaches to determine elasticity values

Regression model	Elasticity formulas
Linear	β ×
Log-log	β
Logistic	β × X × (1 -)

β is the given study’s regression coefficient for a built environmental characteristic. Ȳ is the mean value of running behaviour in that study, while X̄ is the mean value of the built environment characteristic.refers to the mean estimated probability of occurrence

Results

Study characteristics

We identified 39 studies that examined the environmental effects on running activities using VGI data. These studies were published between 2019 and 2024, covering 20 study areas across eight countries on four continents (Fig. 2). Most studies (79.5%, n = 31) focused on Asia, with an intense concentration in China. In contrast, only 10% examined Europe. North America and Oceania were represented in just 5% of studies each (n = 2), while Africa and South America were not. At the city level, Chinese cities dominated, with Beijing (n = 8), Shanghai (n = 6), Chengdu (n = 5), and Guangzhou (n = 4) comprising most studies. By comparison, cities outside China, such as Vancouver, London, Boston, Rotterdam, Helsinki, and Kumamoto (each n = 1), were infrequently studied, revealing a geographic imbalance in the literature.

Fig. 2.

International publication volume over time (A) and by geography (B). Panel B shows individual study locations

Sample sizes varied considerably across the studies, with 33 reporting the number of records in their datasets. These ranged widely, from 40 to 242,265 records, with an average of around 18,000 samples per study. Among the 23 studies that reported volumes of VGI running data, the number of trajectories ranged from 226 to 660,891, with an average of approximately 150,000 per study. This illustrates the substantial scale of available volunteered running data.

Most studies were published in urban studies, environmental science, and public health journals. The three most frequently featured journals were Urban Forestry & Urban Greening, Landscape and Urban Planning, and Frontiers in Public Health. In contrast, fewer studies appeared in journals focused on geography and computer science. Within these fields, only a few journals (e.g., Health & Place, Applied Geography, Computational Urban Science, and Sustainable Cities and Society) featured at least two publications each in this research area.

Definition of runnability outcome variables

The analysis of outcome variables revealed distinct patterns in data sources and methodological approaches. Keep and Strava were the most commonly used sports-tracking platforms (n = 12 and 11, respectively), accounting for 59% of the studies. Edooon followed with nine studies (23.1%), while MapMyFitness (n = 2) and Foooooot (n = 1) were used less frequently (Fig. 3a). In terms of analytical focus, studies were nearly evenly divided: 53.7% focused solely on spatial patterns, while 46.3% analysed both spatial and temporal aspects of running behavior.

Fig. 3.

a) Characteristics in the reviewed studies. Number of sports-tracking apps used and their cross-tabulation with spatiotemporal analysis patterns; b) Number of statistical models used, categorised by type of analysed geographic unit; c) The eight most frequently assessed environmental attributes and the number of studies evaluating each; d) Frequency of different data source types used for specific environmental characteristics

The extracted running activity data demonstrated considerable variability in format, including raw jogging trajectories that offer detailed movement paths [39, 40], aggregated heatmaps that show spatial densities [14, 41], and check-in points that record specific locations [42, 43]. The runnability index was typically measured within predefined buffer zones around runners’ activity locations. Spatial analysis revealed distinct methodological patterns (Fig. 3b). Most studies used street segments as analysed geographic units to assess environmental attributes and runnability (n = 33), while others applied buffers around specific areas such as parks or university campuses (n = 21). Additionally, grid-based units were used in some studies (n = 19). Buffer sizes varied considerably, ranging from 15 m to 1 km. Notably, only one study using street segments incorporated a multi-scale buffer approach, accounting for road hierarchy and varying street widths [44]. For statistical modelling, conventional regression analysis was the most common (n = 31), with spatial regression and machine learning (ML) techniques equally represented (n = 17 each). The total exceeds the article count, as some studies employed multiple analytical approaches.

Our review classified running metrics into seven dimensions: running intensity, frequency, duration, distance, speed, heart rate, and spatial-temporal patterns. Running intensity, also referred to in the literature as “running amount” or “jogging flow”, was the most frequently examined metric (n = 34). It was typically quantified as the total length of jogging trajectories within a given area, normalised by the area’s size. While some studies considered route counts overlapping specific locations as “running frequency” [44–46], we classified these as intensity metrics, as they fundamentally reflect spatial utilisation rates. Running frequency (n = 1) was defined as the number of running activities performed by an individual within a specified time interval, such as weekly or monthly counts [30]. Other trajectory-level performance metrics were less commonly examined, including distance (n = 4), duration (n = 4), speed (n = 3), and heart rate (n = 1). Spatiotemporal analyses revealed distinct geographic patterns in running routes across various time scales, reflecting collective runner preferences and their dynamic associations with the built environment. To ensure a robust and consistent meta-analysis, we focused solely on running intensity as the dependent variable. This was because it was the most frequently examined variable, providing sufficient data for synthesis, while other metrics (e.g., speed, heart rate) were reported in too few studies (n < 3 each) to reliably calculate pooled effect sizes.

Environmental contexts

Natural environment-related variables were the most frequently examined, with blue space provision (59%), Normalized Differnce Vegetation Index (NDVI) (46%), and Green View Index (GVI) (38%) being the most investigated (Fig. 3c). Among building-related and land use factors, land use mix (59%) and points of interest (POI) (51%) were the most studied, followed by building density (44%). Public transport and street density were analysed in 49% of studies for traffic-related factors. In terms of data sources, built environment measures were primarily assessed using remote sensing and satellite imagery, government and public datasets, and map data (Fig. 3d). More recent studies have adopted advanced approaches to improve measurement accuracy, including the integration of high-resolution map data with street view imagery [47, 48] or utilising VGI (e.g., OpenStreetMap) [35].

Associations between running activity and built environment

Built environmental factors were classified based on the consistency of reported findings following Yang et al. [13]. As shown in Table 2, the analysis revealed strong evidence linking features of the natural environment to runnability, particularly for blue space density (38 expected vs. 15 unexpected results), green view index (31 vs. 5), distance to park (19 vs. 7), and park area (27 vs. 1). In contrast, evidence for an association with NDVI was weak. Among built environment characteristics, strong evidence was found for building density (30 vs. 5), floor area ratio (12 vs. 2), and intersection density (27 vs. 8). However, road density (26 vs. 23), POI density (20 vs. 11), and population density (35 vs. 32) showed only preliminary or inconclusive associations.

Table 2.

Strength of associations between environmental attributes and runnability, categorised by expected, unexpected, and null findings

Environmental attributes	Direction of association	Number of studies with expected results	Number of studies with null results	Number of studies with unexpected results
Park area	+	27	0	1
Green view index	+	31	0	5
Building density	-	30	0	5
Floor area ratio	-	12	0	2
Intersection density	+	27	2	6
Blue space density	+	38	2	13
Distance to park	-	19	3	4
POI density	+	20	0	11
Land use mix	-	21	4	16
Road density	+	26	4	19
Public transport density	+	29	0	16
Population density	-	35	3	29
Road connectivity	+	6	0	3
Slope	-	16	2	8
NDVI	+	17	0	24

Dark orange shading indicates strong empirical evidence, while light orange represents emerging evidence. A minus sign (−) denotes that a negative association with running was considered an expected outcome

Elasticity values of built environmental factors

Table 3 presents the number of studies underlying the weighted average elasticity calculations. The analysis identified floor area ratio as the most influential built environment factor, with an elasticity of −0.99. This indicates that a 1% increase in floor area ratio is associated with a 0.99% decrease in runnability. Similarly strong negative associations were found for land use mix (−0.81) and green view index (−0.67), while blue space density showed a substantial positive relationship (0.81). Other environmental attributes had more modest effects, with elasticity magnitudes below 0.3. For instance, park accessibility (−0.01) and public transport density (0.01) had minimal influence, suggesting they may play a limited role in runners’ route selection behaviours.

Table 3.

Results of the weighted average elasticity estimates of environmental characteristics in relation to runnability, and the corresponding number of studies contributing to each estimate

Environmental attributes	Number of available studies	Elasticity value
Floor area ratio	4	−0.99
Land use mix	4	−0.81
Blue space density	6	0.81
Green view index	6	−0.67
Slope	4	0.28
Road density	7	0.13
Park area	3	0.13
NDVI	5	−0.12
Road connectivity	3	−0.11
Intersection density	5	0.04
Building density	6	−0.08
Population density	6	−0.08
POI density	5	0.08
Distance to park	3	−0.01
Public transport density	5	0.01

Discussion

Principal findings

This study systematically reviewed the associations between running behaviors and the built environment, synthesizing evidence from 39 studies. Our meta-analysis revealed strong negative associations between running activity and building density, floor area ratio, and distance to parks. In contrast, strong positive associations were observed with park area, green view index, blue space density, and intersection density. Emerging but less consistent evidence was found for the relationships between running activity and road density, POI density, public transport density, and road connectivity. However, definitive conclusions regarding the association between running and green space (i.e., NDVI) remain challenging due to limited supporting evidence. Our analysis of elasticity values showed that the floor area ratio had the largest effect size, followed by land use mix and blue space density, with distance to parks and public transport density demonstrating the smallest effects.

We found a substantial regional bias, with disproportionate representation from Asia, particularly Chinese cities. Methodologically, Strava and Keep emerged as the dominant tracking platforms, followed by Edooon. Analytical approaches were evenly split between studies focusing exclusively on spatial patterns (approximately half) and those examining both spatial and temporal dimensions of running behavior. Non-spatial regression models predominated in the statistical analyses, though we noted a growing adoption of spatial regression techniques and machine learning (ML) algorithms in recent studies.

Our review shows that research using VGI to examine how the built environment influences jogging has grown rapidly, especially since 2019. This growth reflects advances in geospatial analytics and increasing recognition of fitness-tracking data as a valuable resource for understanding active living behaviours. The literature highlights the potential of sports-tracking platforms to explore spatiotemporal activity patterns, built environment-runnability associations, and population-level health trends. Crucially, these data-driven approaches provide urban planners and policymakers with actionable insights for creating running-friendly environments that support sustainable mobility and public well-being, positioning VGI analytics as a key tool in promoting active lifestyles.

Comparison with previous reviews

Our findings align with a previous review [32] suggesting that runners prefer routes with greater green space exposure than built environments. Green spaces enhance runnability through multiple pathways, including improved connectivity of road network, greater access to blue spaces, and reduced exposure to noise and pollution. Our synthesis of studies provides strong evidence linking runnability to four key environmental factors: blue space density, GVI, floor area ratio, and intersection density. These findings corroborate earlier work by D Huang, B Jiang and L Yuan [20], W Yang, J Hu, Y Liu and W Guo [49], and Y Liu, J Hu, W Yang and C Luo [39]. Notably, our meta-analysis’s first three factors exhibited relatively high elasticity values, indicating a more substantial influence on running behaviour.

The observed negative pooled elasticity for the GVI appears counterintuitive, considering the established preference of runners for green environments such as parks [32]. This finding may be attributable to the scale of analysis employed in the primary studies. In metropolitan-wide analyses, running activity is typically concentrated in denser urban cores characterized by lower green space exposure, whereas extensive peripheral green spaces (e.g., forests, agricultural land) may attract less recreational running. This spatial mismatch can induce a spurious negative correlation at the macro-scale, potentially obscuring positive associations. Furthermore, the relationship is likely non-linear [9, 50]. While moderate green space exposure may encourage running, the effect could plateau or even reverse in areas with very high exposure that lack running-specific infrastructure or are perceived as less accessible or safe [51].

Terrain slope is one of the most debated environmental factors affecting runnability [8, 14]. While traditionally seen as a desirable challenge by distance runners aiming to increase training intensity, our synthesis of crowdsourced data identifies slope as an emerging but negatively correlated factor in runnability assessments. Contrary to conventional assumptions, the analysis suggests that although elite runners may intentionally seek out inclined routes, the broader running population tends to avoid steeper gradients. This pattern is consistently reflected in VGI-derived movement data across multiple urban settings [52–54].

Towards a robust methodological framework

Building on our systematic review, we propose a methodological framework (Fig. 4) that synthesizes the analytical approaches used in VGI-based studies exploring the relationship between the built environment and running activity. Recreational activity-related VGI data has gained increasing attention in studies examining running behavior and built environment associations. Its advantages include high spatial-temporal resolution, cost-effectiveness, large-scale coverage, real-time updates, and a user-generated nature. However, these data also present several limitations, such as inconsistent accuracy, uneven spatial coverage, privacy concerns, and platform bias [26, 33].

Fig. 4.

Methodological framework summarizing the analytical workflow of studies examining environment-running activity relationships using VGI data

Our review identified Strava, Keep, and Edooon as the most popular platforms. The collected VGI data typically falls into three categories: spatial data (e.g., trajectories, or activity-specific check-in points), temporal data (e.g., precise timestamps enabling analysis across multiple time scales from diurnal patterns to annual trends, including pre- and post-event comparisons such as the COVID-19 pandemic), and supplementary data (e.g., runner perceptions, activity details, and text or photo data, which were rarely utilized in the reviewed studies). Built environment data were categorized into four types: natural environment (e.g., green space density, green view index, NDVI), traffic-related features (e.g., street density, public transport density, intersection density), land use characteristics (e.g., land use mix, POI service density), and other factors (e.g., environmental safety, maintenance). Primary data sources included remote sensing, satellite imagery, street view images, government and public datasets, and additional VGI data. These datasets were typically processed using geographic information systems (e.g., ArcGIS) and analyzed through ML techniques (implemented via Python or R).

Our synthesis revealed significant heterogeneity in methodological approaches, particularly in the choice of geographic units, buffer sizes, and statistical models. While this diversity reflects an evolving field, it challenges the comparability of findings and the accumulation of a coherent knowledge base. Below, we critically assess these variations and distil evidence-based recommendations for future research. The geographic unit used to quantify running activity and environmental exposure is a fundamental methodological choice. While grid cells offer simplicity, they poorly match the street network-based nature of running, potentially introducing measurement bias by incorporating non-navigable spaces. Street segments and network buffers are conceptually superior as they align with actual movement paths. Therefore, researchers should prioritize these street network-based assessments over grid cells. The choice between segments and buffers depends on the research question. For example, street segments are ideal for analysing route choice, whereas buffers are more appropriate for assessing exposure to area-based features like land use mix.

Buffer selection critically influences environmental assessments due to the modifiable areal unit problem (MAUP). Although some studies employed multiple buffers to address this issue [41], buffer sizes in the literature varied arbitrarily from 15 m to 1 km. This variation is a primary source of heterogeneity in meta-analyses, as the effect of an environmental variable is inherently scale-dependent. For instance, a 100 m buffer captures immediate streetscape greenery, while a 1 km buffer captures access to a park. To enhance methodological rigor, buffer sizes must be justified theoretically rather than chosen arbitrarily. We recommend a multi-scale approach where feasible. For destination-based features (e.g., parks, transit stations), buffers should reflect reasonable access distances (e.g., 400–800 m network distance). For pervasive features (e.g., general greenness), smaller buffers may be suitable. Researchers should explicitly state the justification for their chosen scale.

The statistical approaches employed also showed a clear evolution. Although conventional regression models were predominant [18, 55], spatial regression and ML methods showed increasing adoption [35, 50, 56]. The field’s reliance on conventional regression prioritizes model interpretability, providing the clear, parametric coefficients essential for informing policy. However, this may mean the evidence base captures only linear effects, potentially missing nuanced threshold effects or complex interactions that ML models can reveal. The choice of model should be driven by the research objective. For policy guidance, conventional or spatial regression remains the standard, as it provides transparent, interpretable effect sizes. For predictive accuracy and exploring complex interactions, ML models are highly recommended, supplemented by interpretation techniques (e.g., SHAP values) to overcome their “black box” nature. Additionally, spatial regression should be routinely considered to validate results, as spatial autocorrelation is prevalent in geographic data.

Future research directions

While objectively measured running intensity dominates current research, subjective runner perceptions warrant greater scholarly attention. Running intensity has emerged as a predominant metric in running-environment studies [9, 14, 55, 57], serving as a key indicator of runners’ route preferences, as evidenced by our review. While this measure effectively captures environmental preferences and route selection behaviours, it offers limited insight into runners’ perceptual experiences during activity, a critical factor influencing participation frequency and long-term engagement [17]. Specifically, subjective dimensions such as running satisfaction [20] and running pleasantness [22], which reflect affective responses to the running environment, remain underrepresented in current scholarship. Recent studies employing GIS methodologies—particularly those utilizing Public Participation GIS (PPGIS) approaches—have begun to address this gap by systematically evaluating these perceptual metrics [58, 59]. These investigations highlight the need to incorporate subjective experiential data beyond purely behavioural measures (e.g., intensity). Social media platforms (e.g., Twitter, Strava Community, or fitness-focused forums) offer rich, untapped sources of self-reported perceptions, sentiment, and qualitative feedback that could effectively complement trajectory analysis [60–62]. By merging these diverse datasets, researchers could develop a more comprehensive framework for assessing built environment-runnability associations, bridging the gap between quantitative metrics and qualitative perspectives.

Current research lacks sufficient focus on individual-level running metrics, with only a limited number of studies examining characteristics such as frequency, duration, distance, speed, and physiological markers (e.g., heart rate) [17, 30, 63, 64]. This narrow analytical scope overlooks critical dimensions of runner–environment associations, particularly how personal running profiles, physiological responses, and psychological restoration needs shape spatial usage [65, 66]. Future research should integrate these individualised metrics with spatial data to develop more comprehensive runnability models that account for both environmental features and runner-specific factors.

Integrating temporal dimensions helps clarify runner–environment dynamics across different timescales. Our review found a notable increase in studies applying spatiotemporal analyses to examine the effects of the built environment on running behaviour [8, 44, 45, 50]. Runnability assessments have started incorporating temporal dimensions with varying analytical depth. While some studies document annual variations [44], others distinguish between weekday-weekend patterns or diurnal fluctuations [45, 50]. Particularly noteworthy are investigations of behavioural shifts during disruptive events like the COVID-19 pandemic [8, 57, 67], which reveal the dynamic nature of these associations. The field exhibits a methodological progression from descriptive temporal analyses (e.g., characterization of running patterns) [45, 49, 67] to explanatory approaches that examine how built environment effects vary across temporal contexts [44, 50]. This evolution highlights the need for dynamic assessment frameworks that systematically integrate short-term variations (e.g., diurnal patterns, weather-related fluctuations), medium-term patterns (e.g., seasonal changes, weekly changes), with long-term shifts (e.g., event-driven, life-course changes). Such temporal integration would advance runnability research beyond static models, enabling more responsive urban planning and policy interventions that account for the dynamic nature of running behaviours.

Our review found that studies primarily operationalized greenspace using simplistic, aggregate metrics such as NDVI or area-based measures [14, 52, 68]. While these approaches capture the presence and visibility of vegetation, they overlook critical morphological qualities like configuration, distribution, and connectivity that fundamentally influence usability. This limitation may explain inconsistent findings in our synthesis, such as the negative elasticity for the GVI. The neglect of greenspace morphology is particularly problematic, as emerging evidence suggests that the shape, arrangement, and internal structure of green features (e.g., tree canopy connectivity, trail network design) are crucial determinants of runners’ route choices [8, 39]. For instance, uninterrupted linear green corridors may facilitate running continuity [17, 20], while clustered vegetation can offer perceptual variety and microclimate benefits [69]. Without accounting for these morphological factors, existing runnability assessments risk oversimplifying environmental quality, potentially leading to urban designs that prioritize green quantity over functional and experiential suitability for runners. Therefore, future research must move beyond aggregate measures to investigate how the greenspace morphology, including shape, connectivity, and internal features, affects its appeal for running. This will require integrating advanced geospatial techniques (e.g., LiDAR-derived 3D metrics, space syntax analysis) with participatory methods to develop metrics that more holistically reflect runners’ needs.

The methodological evolution of this field, as revealed by our synthesis, shows a progression from foundational conventional regressions [16, 17] to GIS-enabled spatial analyses [20, 59]. However, the evidence base remains dominated by linear models, which are likely capturing only main effects and struggling to account for the complex, non-linear relationships emerging in the literature [9, 44, 56]. To better reflect the multifaceted nature of human-environment interactions, a paradigm shift towards advanced analytical techniques is necessary. Machine learning methods, such as random forest, XGBoost, and neural networks, are well-suited to capturing these intricate nonlinearities and interactions that traditional models overlook [56]. A primary obstacle to their adoption has been the “black-box” nature of these algorithms. This challenge can now be addressed through interpretable ML frameworks like SHapley Additive exPlanations (SHAP) [50]. SHAP bridges the gap between predictive power and explanatory insight by providing both global feature importance and local, context-specific explanations, thereby disentangling complex associations [9]. Therefore, we recommend that future runnability studies prioritize the adoption of interpretable machine learning to move beyond static, linear modelling towards dynamic techniques that can more accurately capture the complexity of running behaviours.

Stratified analyses reveal variations in running preferences across different population subgroups, highlighting the need for more tailored approaches to understanding built environment-runnability associations. Our review found that none of the examined studies incorporated stratified analyses based on demographic characteristics (e.g., age or gender). This limitation has obscured essential variations in how different population groups use urban spaces for running. For instance, older adults may prioritise features like smooth surfaces and proximity to green spaces, while gender-related safety concerns often influence route selection among female runners. Incorporating demographic stratification could provide deeper insights into diverse runner needs and support more inclusive urban design [70]. Implementing demographic-stratified analyses offers two key benefits: first, it uncovers subgroup-specific relationships between running behaviours and built environment features; second, it informs the design of more targeted and effective urban planning interventions. Future studies are encouraged to integrate demographic factors to enhance the inclusivity and relevance of runnability research.

Expanding the geographical scope of studies helps reduce regional bias and enhances the generalizability of findings. Our review reveals a pronounced geographical bias in runnability research, with most empirical studies concentrated in China [8, 39, 42, 50]. While this reflects growing interest in this research domain, significant gaps remain in regions facing high environmental health risks, particularly across South and Southeast Asia, the Middle East, Africa, and South America. A more nuanced understanding of how built environments shape running behaviours across diverse global contexts is critical to inform equitable landscape and urban planning strategies, especially in rapidly urbanising areas experiencing dynamic land-use changes. Future research should prioritise broader geographical representation to strengthen the generalizability of findings and adopt open datasets and standardised methodologies to facilitate robust cross-context comparisons.

Cross-platform analysis can improve the generalizability of findings and help mitigate platform-specific biases in runnability assessments. Relying solely on crowdsourced data from a single sports-tracking platform may limit the generalizability of runnability assessments, as such platforms often exhibit inherent demographic and behavioural biases [71]. For example, fitness application users tend to skew toward male, younger, and highly active athletes [33], potentially underrepresenting casual runners, female participants, or older age groups. Furthermore, platform-specific user bases may exhibit distinct recreational activity behaviours [72]. For example, a study conducted in Queensland, Australia, revealed that users of the adventure platform Wikiloc predominantly preferred remote parks with rugged terrain, better representing off-trail users’ behaviours than other platforms [73]. This preference leads to divergent spatial patterns in recorded running intensity across different platforms. Consequently, data from a single source may not accurately reflect the broader population’s usage of street segments, raising concerns about the reliability and representativeness of the findings [74, 75]. To address these limitations, cross-platform analysis is essential. Integrating multiple VGI sources can mitigate platform-specific biases, improve demographic coverage, and yield more robust insights into runnability across diverse user groups. Additionally, VGI datasets should be rigorously evaluated for spatial and demographic population-based representativeness [76], with explicit comparisons to population-level distributions.

Advancing equity metrics in activity space analysis can be achieved by applying population-weighted methodologies, ensuring more representative and inclusive assessments of spatial accessibility and runnability. A key methodological limitation identified in the reviewed VGI-based studies is their failure to incorporate population-weighted approaches when assessing runnability [77]. Current analytical frameworks that disregard population density distributions risk generating biased estimates, as they cannot distinguish whether observed activity concentrations stem from genuinely running-supportive environments or merely reflect spatial artefacts of high residential density [28]. This methodological limitation undermines our capacity to accurately evaluate neighbourhood runnability, particularly in areas where population distribution and built environment quality are misaligned. Integrating population density into runnability indices would significantly improve the equity dimensions of activity space analysis, enabling more precise identification of areas requiring infrastructure improvements to promote physical activity.

Strengths and limitations

This review advances the literature in several ways. First, to our knowledge, this represents the first systematic review incorporating a meta-analysis to examine associations between running activity and built environment characteristics. Second, our work provides novel insights into applying physical activity-related VGI data, highlighting both opportunities and challenges in this emerging methodology. Third, we developed a comprehensive methodological framework that synthesizes and evaluates analytical approaches across VGI-based studies. This framework provides concrete recommendations for future research, including quantification methods for environmental attributes and running metrics, selection criteria for appropriate geographic units and buffer sizes, and guidance for determining suitable statistical modelling approaches.

Several limitations should be acknowledged. First, this review is limited by its focus on crowdsourced GPS data, which excludes insights from surveys or participatory GIS that capture subjective experiences. More critically, the reliance on data from platforms like Strava and Keep introduces significant demographic bias, as they disproportionately represent younger, male, and more athletic users [33]. This underrepresentation of key groups (e.g., older adults, women, lower-income individuals), combined with a lack of stratified analyses in the primary studies, means our aggregated findings may mask important heterogeneity in how different subpopulations respond to built environmental features. Future research must therefore prioritize equitable sampling, conduct stratified analyses by demographics, and develop methods to correct for platform bias to generate universally applicable and equitable insights. Second, the methodological scope of this meta-analysis was constrained by the primary studies, which predominantly employed linear regression models. Consequently, our synthesis captures linear associations but may not reflect potential non-linear relationships between built environment features and jogging behaviour that could be identified using machine learning techniques. Third, our inclusion of GIS and spatial-related terms, which is less common in health literature, may have excluded relevant studies not using such terminology, potentially biasing results toward spatially focused research and underrepresenting broader health-related findings on running and the built environment. Another limitation is the heterogeneity in the measurement of built environment variables across studies (e.g., differing spatial units and buffer sizes). While our findings provide a valuable summary of effects across methodological approaches, future research would benefit from greater methodological standardization to enhance the comparability of results and strengthen evidence synthesis. Fourth, our focus on peer-reviewed journal articles may have omitted findings from grey literature. Finally, the English-language restriction may have introduced geographic bias by excluding potentially relevant studies published in other languages.

Conclusions

This systematic review highlights the transformative potential of VGI and geospatial analytics in understanding how built environments influence jogging behaviours. Our literature synthesis reveals that the field has grown substantially since 2019, driven by the availability of fitness trackers and analytical approaches. Key findings of our meta-analysis demonstrate that factors such as green and blue space density, floor area ratio, intersection density, and terrain slope play critical, yet sometimes contested, roles in shaping runners’ route preferences. Notably, while certain features (e.g., parks, water bodies) consistently promote runnability, others (e.g., slope, high-density urban areas) exhibit more nuanced relationships, reflecting diverse runner needs and preferences. The integration of VGI data offers opportunities for evidence-based urban planning and the creation of running-friendly cities. However, key challenges remain, including data biases, the lack of standardized metrics, and limited consideration of temporal dynamics. Future research should prioritize multi-city comparisons, longitudinal designs, and participatory approaches to support more inclusive and equitable urban environments.

Author contributions

Dengkai Huang contributed to the conceptualization, methodology, data curation, and writing of the original draft. Wenjun Xu performed the formal analysis, visualization, and contributed to the writing of the original draft. Marco Helbich contributed to manuscript review, editing, and supervision. Xiaochun Yang contributed to manuscript review, editing, project administration, and supervision.

Funding

This study was supported by the National Natural Science Foundation of China (No. 52308069), the State Key Laboratory of Subtropical Building and Urban Science (No. 2024ZB16) and Shenzhen University 2035 Program for Excellent Research (NO: 2022B005).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Clinical trial number

Not applicable.

Competing interests

The authors declare no competing interests.