Optimizing Perceived Jogging Supportiveness for Enhanced Sustainable Greenway Design Based on Computer Vision: Implications of the Nonlinear Influence of Perceptual and Physical Characteristics

Abstract

Active transportation and lifestyles are important components of a sustainable city. Greenways play a crucial role in providing conducive environments for jogging. To investigate the influence of micro-scale characteristics on perceived jogging supportiveness (PJS), 230 video clips of greenways within Fuzhou City were collected as samples. PJS was evaluated using a Likert scale, perceptual characteristics were assessed through a semantic difference scale, and physical characteristics were computed via semantic segmentation. By employing SHAP values and dependence plots within an XGBoost framework, the findings reveal the following: (1) Regarding perceptual characteristics, continuity, culture, and facility affordance exhibit the highest relative importance to PJS (|SHAP| ≥ 0.1). Continuity, naturalness, and vitality generally have positive impacts on PJS, while disturbance is negative. Facility affordance, scale, culture, openness, and brightness demonstrate more complex nonlinear influences that suggest optimal value ranges. (2) Concerning physical characteristics, fences, motor vehicles, and surface material are deemed most influential (|SHAP| ≥ 0.1). The presence of fences, walls, and construction generally negatively affect PJS, while excessive openness is also unfavorable. Comfortable road surfaces are associated with higher levels of PJS. Natural elements and the presence of people and vehicles have promoting effects up to certain thresholds, but beyond that point, they exert opposite influences. Finally, suggestions for designing greenways that encourage jogging are proposed. This study provides practical references for optimizing greenway design to promote active transportation and lifestyles, reinforcing the contribution of green infrastructure to public health in sustainable cities.

1. Introduction

Urban greenways play a crucial role in facilitating outdoor jogging activities for urban residents. However, the relationship between the micro-scale characteristics of these greenways and jogging activities remains unexplored, with most studies focusing solely on the macro-scale characteristics and overall physical activity levels. Consequently, the design and optimization of urban greenways to promote jogging can only rely on indirect evidence from studies conducted in other built environments. Moreover, existing research primarily examines objectively measured jogging intensity along routes while neglecting individuals’ subjective perceptions of their surroundings. Furthermore, an analysis framework that combines macro-scale characteristics (overhead-level) with micro-scale characteristics (eye-level) often fails to fully reveal the impact of micro-scale characteristics.

1.1. Greenways and Physical Activity

A sedentary lifestyle and lack of physical activity has emerged as the fourth leading cause of global mortality [1]. Greenways, linear open spaces that offer continuous pathways, present a viable solution for promoting physical activity in various countries and regions worldwide [2]. The existing evidence consistently demonstrates that the establishment of greenways positively influences both moderate-to-vigorous physical activity (MVPA) and overall levels of physical activity (PA) [3,4,5], although a few studies have failed to establish a definitive relationship between greenways and physical activity levels [6,7]. Factors such as community social and economic conditions [8,9], accessibility [10,11], land use mix [9], weather conditions [12,13], and micro-scale design elements [12] serve as moderators in the association between greenways and levels of physical activity.

The current research on greenways and physical activity primarily focuses on the construction and macro-scale characteristics of greenways as causal factors influencing the overall level of physical activity [14]. Some natural experiments have found a significant improvement in the physical activity levels of nearby residents following the construction of greenways [4,8,15]. However, there is limited discussion regarding the relationship between greenways and specific activities. For instance, Hu et al. [16] compared walking levels among residents within 2 km of the East Lake Greenway in Wuhan before and after its construction, while Frank et al. [3] evaluated changes in cycling levels among residents within 300 m of a greenway in Vancouver, Canada, before and after its construction. Additionally, Merom et al. [17] discovered that cyclists were more likely to use greenways, with their construction also promoting cycling activity within their respective areas [18]. Overall, the current research tends to adopt a macro-perspective when examining the relationship between greenways and physical activity; however, knowledge about micro-scale characteristics and specific activities remains lacking.

1.2. Jogging and Built Environment

Jogging is a prevalent physical activity on greenways and is widely recognized as a popular [19], highly health-beneficial [20,21], moderate-to-vigorous physical activity (MVPA). However, the relationship between greenways and jogging remains poorly understood. While some studies have explored the promotion of jogging in various built environments such as urban streetscapes [22,23], large parks [24], and community streets [8], limited attention has been given to the association with greenways. Safety concerns [25,26,27], naturalness factors including vegetation cover and biodiversity presence [20,21,23,24,25,28,29,30,31,32], and continuity [23,26,27,30] aspects along with uninterrupted pathways are frequently discussed in relation to jogging promotion. Additionally, surface materials [25,33,34] and spatial design elements are also considered important factors influencing joggers’ experiences [23,25].

Safety is an essential environmental prerequisite for outdoor joggers. In comparison to indoor joggers, those who engage in outdoor jogging exhibit a heightened perception of potential hazards. The presence of adequate lighting [26,27] and visible safety facilities [25] on urban streets and park paths is positively associated with increased levels of jogging activity and overall satisfaction. Notably, there is a greater prevalence of jogging tracks along relatively open streets [26], as moderate openness facilitates visual connections between the interior and exterior environments, thereby enabling natural surveillance.

Exposure to natural environments has been found to have positive effects on mood regulation and stress reduction, thereby influencing the environmental preferences of joggers [21]. Paths situated in parks and suburban areas are perceived as more appealing compared to urban roads [24,32], particularly those that are closer to water bodies and exhibit a higher presence of vegetation [24,25]. The intensity of jogging [23] and the satisfaction level [30] of joggers have been positively associated with indicators such as the Green View Index (GVI) [29] and the Normalized Difference Vegetation Index (NDVI) [30], as well as the availability and density of blue spaces [20,23,28,29]. While the color green itself does not significantly impact these factors [20], it is the presence of natural elements that truly matters. Both eye-level greenness and top-down greenness are utilized for assessing naturalness levels [28]; however, excessive top-down greenness may impede jogging behavior from occurring effectively [31].

The frequency of jogging interruptions is determined by the continuity of the path. An increased presence of traffic lights [30] and intersections [26,27,30] leads to more interruptions and is associated with decreased jogging intensity on urban streets. When the length of the path exceeds 400 m [23], there is a significant increase in jogging intensity.

In terms of materials, the perceived attractiveness of the environment is closely linked to the comfort of the surface [34]. Joggers exhibit a preference for asphalt paving over gravel, brick, block, and concrete pavements [25,27]. In terms of spatial form, parks with curved paths have higher jogging density but lower intensity compared to straight paths [25]. The relationship between jogging intensity and degrees of sky openness and street enclosure exhibits a nonlinear pattern [23].

In contrast to micro-scale characteristics at the eye level, macro-scale characteristics play a more crucial role in determining the intensity of jogging routes [24,35]. Although some large parks or university sports fields may be suitable for jogging on a micro-scale, urban streets are more accessible and thus more frequently observed as locations for jogging [25,31,36]. Liu et al. [24] found that jogging intensity was lower in areas closer to the city center. Zhong et al. [35] discovered that higher land use density in residential areas led to increased levels of jogging activity compared to those with lower density. Shashank et al.’s [33] development of a city runnability index primarily focused on macro-scale environmental characteristics. As many studies incorporate both macro- and micro-characteristics within the same analytical framework [22,26,30,31], it is essential to fully identify the influence of micro-characteristics.

1.3. Purpose and Questions

The micro-scale characteristics in this research pertain to physical attributes that can be visually observed at eye level, as well as perceptual aspects that are generated through subsequent cognitive processing. In this study, it is hypothesized that certain physical and perceptual characteristics of a greenway play a more significant role in perceived jogging supportiveness (PJS) compared to others, and they exhibit optimal value ranges which can be inferred based on their nonlinear relationships. Consequently, four questions are proposed:

• What perceptual characteristics are relatively significant for the perceived jogging supportiveness (PJS) of greenways?
• How do these perceptual characteristics influence the perceived jogging supportiveness (PJS) of greenways?
• What physical characteristics are relatively significant for the perceived jogging supportiveness (PJS) of greenways?
• How do these physical characteristics influence the perceived jogging supportiveness (PJS) of greenways?

This study aims to comprehensively investigate the influence of greenway characteristics on the perceived jogging supportiveness at a micro-scale. To identify the crucial perceptual and physical characteristic factors, we employ interview surveys alongside semantic segmentation techniques, unveiling the intricate relationship between key characteristic factors and PJS through the application of machine learning techniques (Figure 1), and ultimately recommend optimal value ranges. This study focuses on active travel and healthy lifestyles, which are integral components of sustainable cities, and investigates the influence of the micro-scale characteristics of greenways on the perceived jogging supportiveness (PJS), thereby contributing to a comprehensive understanding of how environmental design impacts individuals’ physical activities. It emphasizes the significance of enhancing jogging experiences by improving both the physical and perceptual aspects of greenways, which is crucial in the fields of urban planning and environmental design. By optimizing greenway designs, greater participation in jogging can be encouraged, leading to enhanced public health and community well-being.

2. Literature Review

2.1. Retrieval Strategy

Currently, there is limited direct research on the influence of greenway environmental characteristics on jogging. Therefore, this study expands the scope of retrieval to include the relevant literature on the impact of built environments on jogging. The search keywords encompass green spaces such as parks and greenways, as well as linear streets and road spaces resembling greenways. Moreover, in order to accurately identify the pertinent literature, the search terms specifically emphasize “physical activity”, “exercise”, and “sport”. The search formula employed in the Web of Science (WOS) database is as follows: TS = (greenway OR park OR street OR road OR “built environment”) AND TS = (run* OR jog*) AND TS = (“physical activity” OR exercise OR sport). The dataset is confined to the Core Collection, drawing from article-type publications in the English language within the last five years from 1 January 2019 to 29 February 2024.

2.2. Literature Screening

In the initial search conducted in the WOS database, a total of 456 articles were retrieved. After eliminating duplicate articles using the NoteExpress literature manager, there remained 452 unique articles. Despite narrowing down the subject area with specific search terms, some of these articles still did not align with the research topic. Firstly, based on their titles, we excluded 156 articles that were not relevant to the subject discipline and an additional 96 articles that did not match our research topic. We also removed four conference papers and four duplicate articles from consideration. This process resulted in a final set of 196 relevant articles. Secondly, by evaluating their abstracts, we excluded 13 papers focusing on non-physical activities, along with another nine unrelated to preference studies and an additional set of 93 that did not investigate environmental characteristics. Furthermore, we eliminated five papers which primarily focused on sports events as subjects. Finally, based on their text, we excluded 50 articles that were not related to environmental perception, six articles that did not classify physical activities, and two articles for which the full text could not be obtained. Consequently, this meticulous screening yielded a final selection of only 18 papers directly related to environmental perception (Figure 2,

Table A1. Extraction table of the literature information on built environment and physical activity.

Literature	Physical Activity	Physical Activity Measure	Environment	Environment Characteristic
				Macroscopic Characteristics	Microscopic Characteristics
					Physical Characteristics	Perceptual Characteristics
Chen et al., 2024 [25]	Walking, running	Activity density, activity intensity	Park	Spatial topology, accessibility	Space form, natural elements, facilities	Aesthetic, safety
Yang et al., 2024 [22]	Running	Jogging flow	Built environment	Building density, intersection density, road density, land use mix, number of parks, number of running tracks, number of water bodies, distance to the nearest subway station, distance to the nearest bus stop, distance to the nearest park, distance to the nearest running track, distance to the nearest water body, Normalized Difference Vegetation Index (NDVI)	Green View Index (GVI), Sky View Index (SVI), Visual Motorization Index (VMI), Visual Humanization Index (VHI), Simpson Diversity Index (SDI)	/
Huang et al., 2023 [28]	Running	Running pleasantness	Built environment	NDVI, blue space availability, intersection density	GVI	/
Huang et al., 2023 [29]	Running	Running intensity	Built environment	NDVI, blue space density, urban density, connectivity, population density	GVI	/
Zhang et al., 2024 [23]	Running	Road running intensity	Built environment	Trail continuity, functional mixing, functional density	Blue space proportion, visual permeability, sky openness, and enclosure	/
Dong et al., 2023 [26]	Running	Running amount	Built environment	/	Buildings, sky, trees, roads, sidewalks	Safety, vitality, beauty, boredom, depression, wealth
Liu et al., 2022 [24]	Jogging	Frequency and pattern of jogging	Park	Accessibility, park area, NDVI, trail density, service facility point of interest density	/	/
Shashank et al., 2022 [33]	Running	Runnability index	Built environment	Street tree density, traffic control facilities, intersection distance, park distance, streetlight density	/	/
Huang et al., 2022 [30]	Running	Perceived satisfaction of runners	Built environment	NDVI, blue space availability, public transport node density, traffic light density, street density	GVI	/
Yang et al., 2022 [37]	Cycling, running	Cycling and running intensity index	Built environment	Land use mix, road density, water area, green space area, riverline length, lighting index, residential building density, building density, floor area ratio, bus and subway station numbers	/	/
Yang et al., 2024 [40]	Jogging	Jogging flow	Built environment	Population density, building density, land use mix, road density, intersection density, distance to the nearest sports facility, number of parks and waterways, distance to the nearest bus stop and subway station.	GVI, SVI, VMI, VHI, SDI	/
Yang et al., 2023 [19]	Jogging	Jogging flow	Built environment	Sky view factors, bus stop density, water feature presence, geographic location, population density, distance to water bodies, distance to parks, distance to bus stops, building density	/	/
Liu et al., 2023 [11]	Jogging	Jogging flow	Built environment	Number of facilities (parks and runways), accessibility of facilities (park, runways, and water bodies), road intersection density, bus stop density, building density, population density, NDVI	GVI, openness	/
Zhong et al., 2022 [35]	Jogging	Relative jogging distance of residents	Built environment	Residential land density (RLD), green land density (GD), arterial road density (ARD), facility diversity (FD), population density (PD), bus stop density (BSD), land use type, road network, building vector, NDVI, natural environment, point of interest (POI)	/	/
Luo et al., 2022 [39]	Cycling, running	Cycling and running activity intensity	Built environment	/	GVI, Sky View Index, Road View Index	/
Zhou et al., 2024 [38]	Running	Running frequency	Built environment	Population density, building density, street density, road intersection density, functional density, POI entropy, street type, bus stop and subway stop density, park green space and greenway density, NDVI	/	/
Deepti Adlakha et al., 2022 [42]	Walking, cycling	The runner’s environmental perception	Greenway	/	/	Walking and cycling paths, lighting, road width and smoothness, bridges and tunnels, traffic risks, separation of cycling and walking paths, natural landscapes, connection to nature
Xu et al., 2022 [43]	Walking, cycling	User perception of greenway health promotion	Greenway	Greenway accessibility	/	Satisfaction and perception, including greenway safety, cleanliness, infrastructure services, and other factors

).

2.3. Macroscopic and Microscopic Characteristics

Research has demonstrated that both macro- and micro-characteristics of the built environment exert a substantial influence on physical activity. Macro-environmental characteristics encompass spatial topology, external traffic accessibility, building density, road density, land use mix, Normalized Difference Vegetation Index (NDVI), and other environmental characteristics. These characteristics contribute to the allure of the environment and subsequently impact jogging flow and intensity. Liu et al. [24] examined the effects of macro-environmental characteristics such as park accessibility, park area, NDVI, trail density, and the density of service facility points of interest on jogging frequency and patterns. Shashank et al. [33] identified that macro-characteristics like street tree density, traffic control facilities, intersection distance, park distance, and streetlight density influence jogging and proposed a runnability index. Yang et al. [37] explored the impacts of macro-environmental characteristics such as land use mix, road density, watershed area, green space area, and residential building density on cycling and jogging intensity. Yang et al. [19] discovered that characteristics like the sky view factor, bus stop density, and presence of water bodies among other macro-environmental characteristics affect jogging flow in built environments. Zhong et al. [35] investigated how macro-environmental characteristics including residential land density (RLD), green land density (GD), arterial road density (ARD), facility diversity (FD), population density (PD), bus stop density (BSD), and land use mix impact relative jogging distances for residents. Zhou et al.’s [38] study revealed that macro-characteristics such as building density, street density, crossroad density, and functional density influence jogging frequency in built environments.

Microscopic environmental characteristics encompass the internal attributes of the environment, comprising the relative proportions of vegetation, sky, roads, and other environmental elements, as well as the distribution and quantity of service facilities. These characteristics are subjectively perceived by users and can significantly influence their visual perception and preferences. However, existing research has relatively neglected microscopic characteristics in this context. For instance, Dong et al. [26] and Luo et al. [39] investigated the impact of microscopic characteristics such as buildings, sky, trees, roads, and sidewalks in street view images on jogging volume and intensity using semantic segmentation.

Previous studies have predominantly employed objective measures to assess environmental characteristics and have examined the combined influence of macro- and micro-characteristics on physical activity and perception. Chen et al.’s [25] research specifically investigated the impact of the density and intensity of walking and jogging activities, revealing that spatial topology and park accessibility are crucial factors influencing jogging at the macro-level. At the micro-level, spatial form, natural elements, facilities, aesthetics, and safety also play a significant role in determining the intensity and density of physical activity. Yang’s two studies focused on analyzing jogging flow within built environments, highlighting the close relationship between building density, intersection density, land use mix, and jogging. Furthermore, micro-environmental characteristics such as the Green View Index (GVI) and Sky View Index (SVI) were found to influence jogging flow [22,40]. Huang’s two studies explored both the pleasantness experienced during jogging in built environments as well as the street segment jog intensity. The findings indicated that the NDVI at the macro-level along with the availability of blue spaces and intersection density were correlated with joggers’ pleasantness as well as their jogging intensity, whereas the GVI significantly influenced both the pleasantness experienced during jogs as well as their intensity [28,29]. Zhang et al. [23] focuses on the intensity of road jogging in urban environments, highlighting the significance of trail continuity, functional mix, and functional density at a macroscopic level. At a microscopic level, characteristics such as the proportion of blue spaces, visual permeability, sky openness, and enclosure also exert substantial influence on jogging intensity. Huang et al. [30], employing objective measurements of the NDVI, blue space density, public transport node density, traffic light density, street density, and micro-level GVI, discovered that these environmental characteristics significantly impact joggers’ perceived satisfaction. Liu et al. [11] investigated the effects of these environmental characteristics on jogging flow by assessing factors including facility quantity, facility accessibility, road intersection density, bus stop density, building density, GVI, and openness at both macroscopic and microscopic levels.

2.4. Physical and Perceptual Characteristics

The perceptual probability model [41] that establishes the framework for the current research on environmental perception suggests that measurable objective physical characteristics serve as indirect cues for subjective emotional responses and behavioral decisions, while perceptual characteristics formed through the cognitive processing of these objective physical characteristics act as direct cues. The physical characteristics of the environment are objectively existing environmental characteristics that can be quantified using scientific methods, typically related to the layout, structure, and composition of the environment. Conversely, the perceptual characteristics of the environment refer to individuals’ subjective feelings and cognition toward it, often encompassing personal evaluations regarding aesthetics, sense of safety, comfort level, and other aspects. Research has demonstrated that both the physical and perceptual characteristics of built environments exert a significant influence on physical activities. Amongst the 18 articles included in this analysis, most studies focused on physically measured characteristics [19,22,24,28,29,30,40], including natural physical characteristics (e.g., GVI, NDVI), architectural environmental characteristics (e.g., enclosure ratio, density), and external traffic-related characteristics (e.g., street density, crossing density). Perceptual environmental characteristics comprise aesthetic perception, safety perception, spatial perception, and environmental comfort level. These perceptual characteristics are not only influenced by the actual conditions of the physical environment but also impacted by individual preferences, cultural backgrounds, and psychological states. Therefore, relying solely on statistical measurements such as jogging flow volume and physical environmental characteristics is insufficient to uncover the direct clues and psychological mechanisms underlying joggers’ decision-making behavior.

Although numerous studies have underscored the significance of perceptual characteristics in elucidating the impact of the environment on jogging, limited research has directly investigated behavioral subjects. Instead, researchers have predominantly relied on open city data or trajectories uploaded by volunteers through smart fitness apps (such as Keep) to objectively measure the jogging of routes [25,26]. However, route jogging flow represents a macroscopic measure, and different constraint mechanisms operate at various scales. Jogging flow is primarily influenced by macroscopic environmental characteristics like accessibility and land use [19,22,40]. The coexistence of macroscopic and microscopic subjective and objective data can potentially obscure the statistical influence of environmental perceptual characteristics. The existing research relatively lacks exploration into subjective user perception. Among the 18 articles retrieved, only a few were based on direct surveys assessing user perception and preferences. For instance, Deepti Adlakha et al. [42] examined pedestrians’ and cyclists’ perceptions regarding greenway environmental characteristics such as road lighting, road width, smoothness, and separation between bicycle lanes and pedestrian paths, as well as natural landscapes, through direct surveys. Xu et al. [43] conducted on-site surveys with pedestrians and cyclists to discuss their perceptions concerning greenway accessibility, satisfaction, and its promotion for health benefits. Therefore, it becomes imperative to separate scales, revert back to a “people-oriented” approach, focus on joggers’ perception of micro-scale characteristics within greenways, directly explore reported environmental needs from behavioral subjects, and provide more direct evidence for design decisions in order to effectively enhance environmental quality from a human perspective and promote public health.

In the academic community, the current research on the relationship between built environments and jogging primarily focuses on urban built environments such as streets and parks, with limited attention given to greenways. Most studies analyze jogging route traffic using objective measures like recorded tracks from fitness apps. However, there is a lack of research that places joggers at the center of the investigation to explore their perception and preferences regarding environmental characteristics. Existing studies often consider both macro-environmental characteristics and micro-environmental characteristics together, without specifically examining the influence of micro-environmental characteristics within the environment. These studies identify environmental characteristics that affect jogging but fail to thoroughly explore how these characteristics impact jogging in a specific manner. Given this, the objective of this study is to address this research gap by focusing on joggers as the research participants and comprehensively examining how the micro-characteristics of urban greenway environments nonlinearly influence joggers’ preferences and willingness to engage, considering both perceptual and physical dimensions.

3. Materials and Methods

3.1. Greenway Videos

3.1.1. Shooting Area

Fuzhou, the capital of Fujian Province, is located in the southeast coastal area of China, downstream of the Minjiang River basin. The terrain is characterized by a basin in the south and mountains in the north, forming a typical estuarine basin landscape. The study area consists of nine selected representative urban greenways within the city of Fuzhou (Figure 3 and Figure 4), such as the Dongjiangbin Greenway, Feifengshan Greenway, Fudao, Guangminggang-Jin’an River Greenway, Huahai Park Greenway, Beijiangbin Greenway, Nanjiangbin Greenway, Wulongjiang Greenway, and the Greenway around Left Sea-West Lake Park (

Table 1. Overview of sample greenways.

Greenway	Surroundings	Length of Selected Line
Dongjiangbin Greenway	River, Estuary, and Street	2.5 km
Feifengshan Greenway	Mountain	1.0 km
Fudao	Mountain	1.2 km
Guangminggang-Jin’an River Greenway	Canal and Street	4.2 km
Huahai Park Greenway	River and Street	3.5 km
Beijiangbin Greenway	River and Street	3.0 km
Nanjiangbin Greenway	River and Street	3.4 km
Wulongjiang Greenway	River, Wetland, and Street	2.0 km
Greenway around Left Sea-West Lake Park	Park and Street	2.0 km

). The aforementioned representative greenways are all situated within the urban area and relatively complete construction. The routes are long enough and are used by a large number of people. These spaces serve as primary venues for Fuzhou residents to partake in leisurely pursuits and recreational activities.

3.1.2. Video Shooting

To replicate the real scene and dynamic visual experiences of joggers in greenways, a GoPro 9 action camera was used to make video recordings of the greenways. Although most joggers typically prefer daytime jogging, the hot and humid climate in Fuzhou due to its unique geographical location makes it unsuitable for engaging in high-intensity physical activities during this time. Therefore, we opted to conduct our shoots early in the morning when temperatures are relatively lower. To ensure their quality, all recordings were made during the day when the weather was good, avoiding periods of direct sunlight and high temperatures. The recording period was concentrated between 7:00 a.m. and 10:00 a.m. in July. The camera was secured to the photographer’s head with a strap. The photographer rode along the center of the road at a steady pace similar to jogging while taking photos, adjusting the height of the bike seat to match the eye level of jogging (Figure 5). A total of 18 videos with a duration of over 15 min each were captured.

3.1.3. Video Editing

The original 18 long videos were first split and trimmed using Boilsoft Video Splitter video editing software (v.8.1.4.0), with a video split duration of 1 min. This resulted in 230 1 min video clips, which were used for the subsequent subjective evaluations of perceptual characteristics and perceived jogging supportiveness. Then, the 230 video clips were further segmented into 1567 images by capturing images every 10 s using the DVDVideosoft Free Studio (v.6.7.2.909) media tool. These images were used for semantic image segmentation to calculate physical characteristics.

3.2. Measures

3.2.1. Perceptual Characteristics and Perceived Jogging Supportiveness (PJS)

Behavior [44,45], aesthetics [27,29], and affection [46,47] are three important dimensions of environmental provision. A greenway well supporting jogging should meet the basic needs of jogging and other outdoor activities, provide high aesthetic value, and create a positive affective experience. Based on the research findings from the previous literature review regarding the impact of built environment characteristics on jogging [24,25,36,48] (Table A1), a comprehensive assessment framework was developed through expert interviews and discussions with jogging enthusiasts. This framework encompasses three evaluation dimensions, namely, behavior, aesthetics, and affection, along with 15 perceptual characteristics that are potentially associated with perceived jogging supportiveness. The behavior dimension includes scale, facility affordance, continuity, and disturbance; the aesthetic dimension includes brightness, color variety, complexity, openness, landscape variability, rhythm, and beauty; and the affection dimension includes safety, naturalness, vitality, and culture (

Table 2. Interpretation and adjective pairs of the perceptual characteristics.

Perceptual Characteristics		Interpretation	Adjective Pair
Dimension	Item
Behavior	Scale	Perceived spatial size	Narrow–Spacious
	Facility affordance	Environmental design and facilities to meet the basic needs of outdoor activities	Inconvenient–Convenient
	Continuity	Degree of difficulty between the start and end points of the path	Discontinuous–Continuous
	Disturbance	The number of factors in the environment that may disrupt the rhythm of jogging or even interrupt the jogging	Little Interruption–Much Interruption
Aesthetics	Brightness	Perceived rightness of the environment	Dim–Bright
	Color variety	The variety of hues present in the environment	Colorless–Colorful
	Complexity	The complexity of cognitive processing required for visual landscape information in the environment	Monotonous–Complex
	Openness	The degree of openness of the space	Enclosed–Opened
	Landscape variability	The changing perspective effects obtained with the movement of the viewpoint	Little Change–Much Change
	Rhythm	The perception of the temporal rhythm of the appearance of objects with the movement of the viewpoint	Weak Rhythm–Strong Rhythm
	Beauty	The aesthetic quality of the landscape	Lack of Aesthetics–Rich in Aesthetics
Affection	Safety	Feelings of personal and property safety in the environment, without fear or feeling threatened	Dangerous–Safe
	Naturalness	The degree of perception of closeness to nature in the environment	Artificial–Natural
	Vitality	The environment is vibrant, dynamic, and attractive, capable of stimulating people’s interest and participation	Lethargic–Energetic
	Culture	The degree of cultural value and perceived significance in the environment	Uncultured–Cultured

). Perceptual characteristics were measured using the semantic differential method, transforming 15 perceptual characteristics into 15 pairs of adjectives (Table 2), with the adjective on the far left being assigned a score of −2 and the adjective on the far right being assigned a score of 2, totaling five rating levels.

The evaluation of PJS included two questions, namely, “To what extent do you think this environment supports jogging activities?” and “To what extent are you willing to jog in this environment?”. The evaluation used a 5-point Likert scale, with higher numbers representing higher scores. The average of the scores for the two questions is used as the final value for PJS.

In order to ensure the rigor of the evaluation process and the uniformity of evaluation criteria, we invited 34 college students who are avid joggers to participate in assessing perceptual characteristics and their perceived jogging supportiveness in greenway videos. To enhance objectivity and minimize mutual influence, a “back-to-back rating” method was employed, where two raters independently rated each perceptual characteristic. Consensus was reached through in-depth discussions when there were discrepancies in the ratings, ensuring the accuracy and reliability of evaluations. The final evaluation result for each characteristic was determined by averaging the ratings from both raters when they were consistent [49,50]. This approach not only improved the consistency but also enhanced the credibility of the research findings. Each rater was required to evaluate all 230 video samples, and before making the evaluations, they underwent training to ensure they understood the meaning and method of the items being rated (Tables S1 and S3). The videos were played on a laptop with a 15.6-inch screen and a resolution of 3840 × 2160 pixels.

3.2.2. Physical Characteristics

Semantic segmentation is a computer vision technology that uses computers to identify and understand the content of images at the pixel level, effectively extracting landscape elements from the images [51]. This study selected the Pyramid Scene Parsing Network (PSPNet) for semantic segmentation modeling and used the ADE20K public dataset to pre-train the PSPNet [52]. The ADE20K dataset covers a variety of scenarios such as indoor, outdoor, and natural scenes. It is a large dataset with over 25,000 densely annotated images, 150 categories of object labels, and 3000 specifically annotated objects [53].

In PyCharm software (v.2024.1.3.0), semantic segmentation was performed on a dataset of 1567 images extracted from 230 video clips. Due to the large number of labels recognized by ADE20K and some labels being irrelevant to outdoor spaces, it was necessary to filter out 150 classes of labels, retaining only those related to greenway landscapes and facilities while deleting indoor facility and furniture labels. Then, the filtered labels were classified and merged; for example, all plant labels recognized by semantic segmentation were merged into the Green View Index (GVI), while all water labels were merged into the Blue View Index (BVI). Finally, physical characteristics that may affect the same perceptual characteristics were combined into composite indicators. GVI, BVI, and animal-related characteristics were combined with natural elements to create the Natural View Index (NVI). Based on the segmentation results, 20 physical characteristics were calculated (

Table 3. Interpretation and formulas of the physical characteristic indicators.

Physical Characteristic Indicators	Interpretation	Formula
Green View Index (GVI)	Pixel proportion of vegetation greening in the image	$GVI={\displaystyle \frac{A_{tree}+A_{grass}+A_{plant}+A_{flower}+A_{palm}}{A_{total}}}$
Blue View Index (BVI)	Pixel proportion of water bodies in the image	$BVI={\displaystyle \frac{A_{water}+A_{sea}+A_{river}+A_{lake}}{A_{total}}}$
Nature View Index (NVI)	Pixel proportion of natural elements (plants, water) in the image	$NVI={\displaystyle \frac{GVI+BVI+Animal}{A_{total}}}$
Enclosure Index (EI)	Degree of enclosure of the greenway space in the image	$EI={\displaystyle \frac{A_{building}+A_{wall}+A_{fence}}{A_{total}}}$
Landscape Variety Index (LVI)	The negative sum of the pixel ratio of each landscape element multiplied by the natural logarithm of its value	$LVI=-\sum P_i\ln P_i$
Sky	Pixel proportion of sky in the image	$Sky={\displaystyle \frac{A_{sky}}{A_{total}}}$
Animal	Pixel proportion of animals in the image	$Animal={\displaystyle \frac{A_{animal}}{A_{total}}}$
People	Pixel proportion of people in the image	$People={\displaystyle \frac{A_{people}}{A_{total}}}$
Culture facility	Pixel proportion of cultural facilities (information boards, cultural walls, sculptures, landscape stones) in the image	$Culturefacility={\displaystyle \frac{A_{rock}+A_{signboard}+A_{fountain}+A_{sculpture}+A_{flag}}{A_{total}}}$
Service facility	Pixel proportion of service facilities (benches, trash bins) in the image	$Servicefacility={\displaystyle \frac{A_{chair}+A_{bench}+A_{ashcan}}{A_{total}}}$
Hard surface	Pixel proportion of roads and squares in the image	$\begin{array}{l}Hardsurface\\ ={\displaystyle \frac{A_{floor}+A_{road}+A_{sidewalk}+A_{earth}+A_{path}+A_{runway}+A_{bridge}+A_{land}}{A_{total}}}\end{array}$
Stairs	Pixel proportion of steps in the greenway path in the image	$Stairs={\displaystyle \frac{A_{stairs}+A_{stairway}}{A_{total}}}$
Motor vehicle	Pixel proportion of motor vehicles in the image	$Motorvehicle={\displaystyle \frac{A_{car}+A_{bus}}{A_{total}}}$
Non-motor vehicle	Pixel proportion of non-motor vehicles in the image	$Non-MotorVehicle={\displaystyle \frac{A_{minibike}+A_{bicycle}}{A_{total}}}$
Walls	Pixel proportion of walls in the image	$Wall={\displaystyle \frac{A_{wall}}{A_{total}}}$
Construction	Pixel proportion of buildings in the image	$Construction={\displaystyle \frac{A_{building}+A_{house}}{A_{total}}}$
Fences	Pixel proportion of fences in the image	$Fence={\displaystyle \frac{A_{fence}+A_{railing}+A_{bannister}}{A_{total}}}$
Streetlight	Pixel proportion of streetlights in the image	$Streetlight={\displaystyle \frac{A_{streetlight}}{A_{total}}}$
Monitor	Pixel proportion of surveillance facilities in the image	$Monitor={\displaystyle \frac{A_{monitor}}{A_{total}}}$
Surface material	Material for laying road surface	Brick paving = −2; block paving = −1; concrete paving = 0; asphalt paving = 1; plastic paving = 2.

Among them, $A_{total}$ represents the total number of pixels in the image, and $A_x$ represents the number of pixels of a certain element in the image; $P_i$ represents the pixel ratio of a certain element in the image.

and Figure 6). Since the semantic segmentation model cannot identify pavement material, manual scoring was applied. The plastic paving was assigned 2 points, asphalt 1 point, concrete 0 points, block −1 point, and brick −2 points (Tables S1 and S2).

3.3. Analytical Methods

3.3.1. Xtreme Gradient Boosting

Xtreme Gradient Boosting (XGBoost), an efficient ensemble learning algorithm developed by Tianqi Chen based on the boosting framework [54], is an improvement of the gradient-boosting decision tree (GBDT). Compared to the GBDT, XGBoost performs a second-order Taylor expansion of the loss function and adds L1 and L2 regularization terms, achieving overall optimality while preventing overfitting [55]. In this study, we utilized the “xgboost” package in the R 4.3.2 platform to link PJS with latent factors at both the physical and perceptual levels. The datasets were divided into 80% for the training set and 20% for the test set. Since hyperparameters can affect a model’s fitting performance, five-fold cross-validation was employed to determine the optimal hyperparameters. The accuracy of the adjusted models was then evaluated using the root mean square error (RMSE) and the coefficient of determination (R²). RMSE is calculated by first finding the mean of the squared prediction errors for all samples (MSE) and then taking the square root of this mean. A lower RMSE indicates higher model accuracy. R², on the other hand, represents the proportion of the total variance that is explained by the model. It ranges from 0 to 1, with values closer to 1 indicating a better fit of the model.

3.3.2. Shapley Additive Explanation

Machine learning models, often regarded as “black boxes”, are more difficult to interpret than traditional models (e.g., linear regression models). However, the interpretability of these models can be enhanced by using Shapley additive explanation (SHAP), a tool that overcomes this limitation. SHAP is an extension of the Shapley value concept from game theory and is primarily used to quantify the contribution of each characteristic to the model’s predictions [56]. It calculates the marginal contribution of a characteristic when added to the model and then computes the marginal contributions of that characteristic across all possible characteristic sequences, ultimately yielding the SHAP values for that characteristic [57].

Assume the ith sample is denoted as $x_i$, and the jth characteristic of the sample is $x_{jj}$. The model’s prediction for this sample is $y_i$, and the baseline of the model (usually the mean of the target variable across all samples) is $y_{base}$. The calculation of the SHAP values is shown in Equation (1):

$$y_i=y_{base}+f\left(x_{i1}\right)+f\left(x_{i2}\right)+f\left(x_{i3}\right)+\dots ..+f\left(x_{ij}\right),$$

(1)

$f\left(x_{i1}\right)$ represents the contribution of the first characteristic of the ith sample to the final predicted value. The SHAP values for each characteristic indicate the expected change in the model’s prediction when conditioned on that characteristic. If $f\left(x_{i1}\right)$ > 0, it means that the characteristic increases the predicted value. Conversely, if $f\left(x_{i1}\right)$ < 0, it means the characteristic decreases the contribution to the prediction.

In this study, we utilized the “shapviz” package in the R 4.3.2 platform to interpret the XGBoost model. This allowed us to generate characteristic importance ranking plots, SHAP summary plots, and dependency plots, which help explain the contribution of each characteristic to the predictions and the relationships between characteristic values and model output (Figure 5).

4. Results

4.1. Hyperparameter Tuning of XGBoost Models

The XGBoost model was tuned using a combination of hyperparameters (

Table 4. XGBoost hyperparameter selection.

Hyperparameter	Explanation	Range of Values	Adjusted Value for Perceptual Dimension	Adjusted Value for Physical Dimension
Learning_rate	Shrinkage step size in the learning update process	[0.01, 0.5]	0.04	0.01
Max_depth	Maximum depth of the tree	[1, 10]	9	10
Colsample_bytree	Percentage of columns per random sample	[0.2, 1.0]	0.4	0.4
Min_child_weight	Sum of sample weights of minimum leaf nodes	[1, 10]	2	1
subsample	Percentage of random samples per tree	[0.2, 1.0]	0.7	0.4
Gamma	Leaf node splitting threshold	[0.1, 3]	0.1	0.2

). The model adjusted for the perceptual dimension achieves a root mean square error (RMSE) of 0.599 and a coefficient of determination (R²) of 0.633, while the model adjusted for the physical dimension has an RMSE of 0.566 and an R² of 0.616. Drawing on previous research, both models have R² values exceeding 0.6, indicating a good fit. Although RMSE comparability across different studies is limited due to variations in data scales, the RMSE in this study remains within a reasonable range. Overall, both models exhibit satisfactory fitting performance [58,59,60].

4.2. The Influence of Perceptual Characteristics on PJS

4.2.1. Relative Importance

Figure 7a illustrates the relative importance ranking of perceptual characteristics. The mean absolute value of SHAP decreases from top to bottom, and the three different colors represent the three subdimensions of perceptual characteristics. The relative importance ranking of the subdimensions, from highest to lowest, is behavior (40.0%) > aesthetics (30.2%) > affection (29.8%), with behavior having the greatest influence, while aesthetics and affection are nearly equal. Among the specific perceptual characteristics, the top five are continuity, naturalness, culture, facility affordance, and scale. Continuity has the strongest influence on PJS, and naturalness is significantly stronger than all other characteristics except continuity. This highlights that spatial continuity and the natural feel of the environment play an important role in influencing PJS. Notably, the strongest characteristic within the aesthetic dimension—openness—ranks sixth, indicating that at the three subdimensions within the perceptual characteristics, aesthetics is relatively less important compared to behavior and affection.

4.2.2. SHAP Summary Plots

Figure 7b illustrates the positive or negative impacts of each perceptual characteristic on PJS. The color of the points represents the magnitude of the original feature values, with darker colors indicating lower feature values and lighter colors indicating higher feature values. Continuity, naturalness, and vitality show positive impacts, demonstrating that improving spatial continuity, the natural feel of the environment, and vitality can effectively enhance PJS. In contrast, openness and brightness exhibit negative impacts, indicating that overly open and bright environments have a strong inhibitory effect on PJS.

4.2.3. SHAP Dependence Plots

Since the collective relative importance of the first nine perceptual characteristics (continuity, naturalness, culture, facility affordance, scale, openness, brightness, vitality, and disturbance) accounts for 81.8%, we chose these variables to create SHAP dependence plots. Figure 8 presents the SHAP dependence plots with LOWESS (locally weighted scatterplot smoothing) fitting curves, which better illustrate the relationships between these characteristics and their SHAP values.

The SHAP values show a positive correlation with continuity and vitality, though the patterns of change differ slightly. SHAP values increase as continuity rises, with a threshold of around 0, where SHAP values shift from negative to positive; beyond this point, the rate of increase slows down. For vitality, the line shows a more consistent upward trend, with the SHAP value becoming positive at a threshold of 1. This indicates that higher spatial continuity and environmental vitality strengthen the positive influence on PJS. Naturalness generally exhibits a positive correlation with SHAP values but declines between 0 and 1, only becoming positive at a value of 2. This implies that naturalness must reach a certain level to exert a beneficial influence on PJS.

The relationships between the SHAP values and culture, facility affordance, scale, and disturbance follow an “N-shaped” pattern, showing an initial rise followed by a decrease. These factors show positive SHAP values only within specific ranges (culture: 0–3; facility affordance: −1–1; scale: −1–1; disturbance: −2–0), and their strongest positive impacts occur at particular values (culture: 1; facility affordance: 0; scale: 0; disturbance: −1). This indicates that moderate levels of cultural attributes, facility availability, environmental scale, and disturbance factors are conducive to PJS.

Openness and brightness also display an “N-shaped” trend in relation to the SHAP values but with differing rates of increase and decrease. Both characteristics show specific ranges where they positively influence PJS, though their overall effect is relatively limited. However, excessively high levels of environmental openness and brightness significantly diminish PJS.

4.3. The Influence of Physical Characteristics on PJS

4.3.1. Relative Importance

Figure 9a illustrates the relative importance ranking of physical characteristics. The top five characteristics and their respective proportions are fences (15.9%), motor vehicles (14.3%), surface material (11.2%), Enclosure Index (10.0%), and Sky (9.3%). Notably, the first three characteristics account for more than one-third of the total importance, indicating that fences, motor vehicles, and surface material in the environment significantly influence PJS. In contrast, natural factors such as the NVI and GVI have relatively lower importance compared to the aforementioned artificial factors, suggesting that natural elements have a lesser impact compared to some artificial elements. The relative importance of monitoring and animals is extremely low, indicating that characteristics like environmental safety and small animals such as birds commonly found in parks do not significantly affect PJS.

4.3.2. SHAP Summary Plots

Figure 9b illustrates the positive or negative impacts of each variable within the physical characteristics on PJS. Surface material and the Green Visual Index (GVI) generally show a positive impact, indicating that improving the comfort of surface material and increasing the GVI can enhance PJS. In contrast, the EI and Sky exhibit a negative impact overall, suggesting that excessively high levels of the EI and sky visibility are not conducive to jogging activities.

4.3.3. SHAP Dependence Plots

Since the collective relative importance of the first twelve physical characteristics (fences, motor vehicles, surface material, EI, Sky, people, GVI, non-motor vehicles, construction, NVI, walls, and stairs) reaches 82.2%, we selected these variables to create SHAP dependence plots with LOWESS fitted curves (Figure 10).

The trends of the SHAP values for fences, motor vehicles, non-motor vehicles, and stairs initially show a rapid decline before stabilizing or slightly recovering. Specifically, the threshold where the SHAP values for fences become negative is around 0.01, whereas for motor vehicles, it is 0.002. For non-motor vehicles and stairs, most points fall below the SHAP value of 0, indicating an overall adverse effect. Additionally, since most points for motor vehicles, non-motor vehicles, and stairs are clustered around low values, the upward trend observed later in the figures may be less reliable. Generally, fewer fences, motor vehicles, non-motor vehicles, and stairs are more favorable for jogging.

The relationship between the SHAP values and surface material, Sky, and construction follow an approximately monotonic trend. As the quality of the surface material improves, the SHAP values increase, with a threshold at 1. This indicates that more comfortable surface materials are more beneficial for jogging, while poor-quality materials have an inhibitory effect. Conversely, the relationships between the SHAP values and Sky and construction are monotonic decreases, with thresholds at 0.08 and 0.03, respectively. This suggests that while a lower sky visibility and building proportions initially enhance PJS, exceeding these thresholds leads to negative effects.

The relationship between the SHAP values and people, GVI, and NVI shows an initial increase followed by a decrease. For people, the SHAP values rise rapidly at first and then decline slowly, with a positive influence observed after 0.002, peaking around 0.008. This indicates that encountering an optimal number of people during jogging is beneficial. The trends for the GVI and NVI are more gradual, with positive effects seen between 0.5 and 0.85 and between 0.7 and 0.9, respectively, peaking around 0.65 and 0.75, respectively. This implies that higher and well-balanced levels of the GVI and plant elements are needed to enhance PJS.

The nonlinear relationships between the SHAP values and the EI and walls are more complex. For the EI, the SHAP values remain relatively stable until around 0.1, after which they begin to decrease, becoming negative around 0.12, with a slight recovery at 0.2, though they remain negative. Walls exhibit a less pronounced positive impact, initially decreasing and then fluctuating around 0 between 0.01 and 0.05, with a direct decline after 0.05. Both the EI and walls have optimal ranges: the EI positively promotes jogging as long as it is below 0.1, while walls do not negatively impact jogging if kept below 0.05.

5. Discussion

5.1. Perceptual Characteristics’ Influence on the PJS of Greenways and Implications for Design

5.1.1. Behavior Dimension

Perceptual characteristics related to behavior exert the most significant influence on PJS. To maintain a steady jogging rhythm, it is essential to have continuous paths with spatial scales coordinated for jogging and minimal disturbances. Adequate facility affordance ensures the fundamental requirements for outdoor activities. Typically, daily jogging sessions last more than 20 min and cover a distance of over 2 km [61]. A sufficient path length with fewer interruptions enables joggers to sustain a balanced rhythm, prevent sports injuries, and enjoy an optimal exercise experience. Previous studies indicate that excessive traffic breakpoints and infrastructure are associated with reduced jogging traffic [36], while a minimum path length of 400 m is necessary for achieving satisfactory levels of jogging experience [23,61]. The study conducted by Chen et al. revealed that the continuity of trails exerts a positive influence on activity density [25]. Uninterrupted trails can minimize disruptions in walking and jogging, thereby enhancing the exercise experience. These findings align with our own research outcomes, as continuous jogging paths are frequently associated with an optimal jogging encounter. In terms of design considerations, greenway route selection should avoid urban core areas characterized by high road network density in order to prevent disruptions caused by traffic breakpoints. If unavoidable, connecting facilities such as overpass bridges or underground passages can be considered. Furthermore, apart from ensuring horizontal continuity within the space layout, attention should also be given to avoiding steep slopes and excessive stairs in the vertical direction.

Regarding facility affordance, the absence of fundamental amenities like restrooms and kiosks can impede meeting basic needs during outdoor activities. Conversely, an excessive provision of service facilities offers visitors additional opportunities to pause and engage in diverse activities such as sitting, conversing, or playing musical instruments. However, this may result in crowds gathering, thereby disrupting the jogging rhythm, reducing jogging speed, and diminishing the intensity of jogging [62]. The conclusion of this study aligns with our research findings, which suggest that inadequate service facilities fail to meet the fundamental needs of users and consequently result in a diminished level of PJS. Conversely, an excessive provision of service facilities can impede operations and thereby hinder PJS. Consequently, the number of service facilities along the greenway should align with visitor capacity, while their spatial arrangement should be strategically positioned away from the jogging route to prevent creating large stopping areas that could lead to crowd formation and interrupt jogging.

In terms of scale, there exists a typical inverted “U-shaped” relationship between scale and the SHAP values. The results of previous studies align with our findings, suggesting that activity density is influenced by spatial scale. Both excessively small and large spatial scales may impede walking and jogging activities [24,25]. Various speeds and intensities of physical activity correspond to different sizes of buffer spaces, with insufficient buffer spaces resulting in collisions. Generally speaking, higher speeds require larger buffer spaces, and joggers necessitate more space compared to walkers [26]. Excessively large or wide spaces are not conducive to route fixation and crowd decentralization; they can also evoke negative emotions. In the design and maintenance of greenways, for paths that are too narrow, appropriate spatial concessions should be made by landscape elements on both sides; plants with low branching points ought to be reduced, while regular trimming is necessary to prevent collisions. For paths that are excessively wide, secondary spatial division can be guided by spray-painted dividing lines for directing routes or increasing planting density on both sides to enhance the sense of enclosure.

In terms of disturbance, excessive interference factors can disrupt the jogger’s exercise process, thereby impeding jogging performance. Jogging is a high-intensity physical activity compared to walking, and excessive disruption can not only impede the jogger’s speed but also pose physical risks. Previous studies have also shown that people and vehicles are significant sources of disturbance during jogging [27,36], with an excess number negatively impacting the jogger’s pace. Previous studies have identified people and various types of traffic as disruptive factors for joggers [63]. Introducing a limited number of disturbances appropriately increases the challenge of jogging while enhancing environmental liveliness. When designing greenways, it is crucial to prohibit large vehicles such as motor vehicles from entering to prevent potential traffic accidents. Additionally, separate exercise routes should be established for non-motorized vehicles and pedestrians in order to segregate jogging tracks from walking paths and cycling lanes.

5.1.2. Affection Dimension

Greenways that possess a cultural ambiance and exhibit vibrancy and naturalness are more likely to be embraced by joggers. Among perceptual attributes, naturalness exerts a significant influence on perceived jogging satisfaction (PJS), second only to continuity. The Stress Recovery Theory (SRT) proposed by Ulrich et al. posits that exposure to green spaces can mitigate negative emotions such as stress and enhance emotional well-being [64]. Natural exposure is one of the contemporary residents’ requirements for utilizing green public spaces, necessitating less artificial modification and more natural landscapes in such areas. Most joggers prefer jogging on greenways that can simultaneously fulfill their need for connecting with nature and engaging in physical activity, thereby endowing greenways with more natural landscapes with higher PJS scores. This finding aligns with prior research indicating that jogging satisfaction positively correlates with natural exposure [24,30]. When selecting greenway routes, it is advisable to avoid bustling downtown areas or high-traffic zones as excessive traffic noise may impede the desired natural experience. If the route traverses rivers, lakes, or other water features, it should leverage these elements through techniques like leaving blank spaces. Regarding vegetation selection, native plants and trees should be prioritized while increasing plant density moderately without overdoing it since excessive vegetation could attract mosquitoes or create blind spots leading to suboptimal jogging experiences and reduced environmental safety.

In terms of culture, a vibrant cultural ambiance can enhance the appeal of greenways, with cultural influence ranking third among all perceptual characteristics. Relevant studies have also indicated that joggers take cultural factors into account when selecting their routes [27,33], aligning with the findings of our research. A moderate number of cultural amenities, such as informative signboards, landscape walls, and sculptures, can stimulate interest in jogging and encourage joggers to explore the landscape design further. However, an excessive abundance of cultural facilities may divert joggers’ attention and disrupt their jogging experience [62]. During the design phase, it is advisable to judiciously increase the presence of cultural facilities near the greenway. For instance, sculptures could be strategically placed on spacious lawns as focal points, while landscape stones could be utilized for embellishment alongside plant groupings. Additionally, information boards or cultural walls could be arranged in squares to promote local cultural attributes.

For vitality, a dynamic greenway environment can offer joggers heightened levels of engagement and bolster their motivation to engage in jogging activities. Previous studies have also suggested that joggers consider a sense of vitality as one of the factors when selecting routes, since routes with such a feeling are more attractive to joggers and thus encourage PJS [26,27,36]. Greenway environments characterized by untidy vegetation, limited human presence, monotonous color schemes, and excessively linear paths lack a sense of liveliness. Neglected vegetation may result in reduced usage, monotonous colors can induce visual fatigue among joggers, and overly straight paths can make the jogging experience too predictable and less captivating [25]. During the design phase, enhancing color variation can be achieved by enriching the hue of the jogging track and selecting plants with diverse shades for planting. Additionally, route design should incorporate a combination of straight and curved paths to avoid monotony in path formation.

The impact of safety on PJS is found to be the lowest in this study, which contradicts previous research and practical experience; previous research has demonstrated that the provision of a secure jogging environment can foster increased engagement in jogging and enhance the contentment and convenience experienced by joggers [25,26,27,33,36]. This discrepancy may arise from the evaluators’ reliance on video assessments, which lack an immersive experience. Additionally, the greenway videos were recorded during daylight hours, potentially enhancing perceived safety. Consequently, our findings suggest a relatively limited influence of safety on PJS.

5.1.3. Aesthetics Dimension

A moderate level of openness can enhance safety and establish a visual connection between the interior and exterior, thereby promoting an optimal environment for joggers. A study suggests that an enhanced street environment, characterized by abundant natural scenery and a reduced sense of confinement, may potentially foster jogging engagement by increasing its appeal [26]. Conversely, excessive openness fails to provide adequate shelter and seclusion for joggers. Yang and Fei also discovered that a balanced degree of openness is advantageous for joggers [22]. Therefore, in excessively open environments, the sense of enclosure can be heightened by strategically planting tall trees and shrubs while leveraging changes in terrain to optimize vertical design.

Regarding brightness, an excessively dim environment can impede joggers’ clear perception of their surroundings and route, inducing a sense of unease toward the external environment and consequently impacting the jogging experience [65]. Appropriately augmenting brightness within a specific range not only enhances aesthetic appeal but also promotes jogger safety [26,27,36]. An overly bright setting is often accompanied by intense sunlight, causing dizziness and visual discomfort for joggers. Therefore, the strategic placement of vegetation can be employed to optimize brightness levels. Tall broad-leaved trees offer shade to mitigate direct sunlight discomfort while considering vertical plant layering that avoids excessive density to prevent the complete obstruction of external light.

Given that jogging involves dynamic movement, joggers’ perception of the scenery also undergoes changes as their viewpoint shifts. Therefore, it is crucial to pay meticulous attention to dynamic indicators such as rhythm, continuity, variability, and disturbance in the design process. By establishing a well-balanced rhythm for the appearance of scenery along the greenway, its overall rhythmic quality can be enhanced. Furthermore, modifying various aspects of the environment including space, terrain, structures, plants, and textures can contribute to increased variability along the greenway.

5.2. Physical Characteristics’ Influence on the PJS of Greenways and Implications for Design

5.2.1. Fences, Walls, and Constructions

Fewer artificial built factors such as fences, walls, and constructions can promote PJS. Previous research has demonstrated that the physical characteristics of street enclosures can impact joggers’ perceptions of safety, comfort, and privacy, consequently influencing their engagement in jogging [26]. The impact of fences on PJS is strong and has a threshold. A suitable number of fences is necessary. In some areas close to water or when passing on bridges, railings play a protective role in preventing users from falling, which improves PJS. However, too many fences can create a sense of restriction, thereby reducing the jogger’s satisfaction [22]. In non-waterfront areas, it is recommended to use low shrubs or hedges instead of artificial fences to provide protection while avoiding a strong sense of restriction.

When the proportion of walls is greater than 0.05, it can block the jogger’s line of sight and sometimes interfere with the jogging movement, thereby inhibiting PJS [22]. When the proportion is less than 0.05, it may promote PJS because walls that are far from the greenway do not affect the jogging experience. Therefore, an open design approach can be adopted to seamlessly connect the greenway with surrounding parks, squares, streets, and other public spaces, forming a coherent network of public spaces. Landscape elements that are harmonious with the surrounding environment, such as sculptures, benches, and lighting fixtures, can also be used to replace walls, increasing the greenway’s affinity and aesthetics.

When the proportion of construction is higher than 0.03, it means that the construction is too large or too close to the road, which can interfere with jogging behavior and inhibit PJS [35]. Construction with a visual ratio of less than 0.03 can serve as a means of orientation for joggers or as embellishments to the environmental landscape, thereby promoting PJS. The selection of greenway routes should avoid areas near tall buildings, as this can not only block the jogger’s line of sight but also cause a sense of oppression.

5.2.2. People, Motor Vehicles, and Non-Motor Vehicles

When exceeding a certain number, people, motor vehicles, and non-motor vehicles may become disturbances for jogging. The impact of motor vehicles on PJS is second only to fences. As the proportion of motor vehicles increases, they not only affect the jogging process but also pose a threat to the personal safety of joggers, thereby affecting the jogging experience. Schuurman et al. found that 61% of respondents would worry about being attacked by motor vehicles during their jog [27,66]. In the design of greenways, motor vehicles should be prohibited from entering the greenway. Physical barriers such as isolated green belts should be set up to ensure a clear separation between the greenway and motor vehicle lanes.

For non-motor vehicles, jogging and cycling are usually allowed in the same route, raising the risk of collision between different exercises and thus hindering the willingness of jogging [66]. In consideration of safety, it is better to set independent routes for jogging and cycling or separate spaces for different exercises using ground makers or planted barriers.

For the factor of people, if there are no people or too few people on the greenway, it may be perceived as desolate, lacking in vitality, leading joggers to feel unsafe or threatened [30]. However, when too many people are present, it can lead to crowding and interrupt jogging. Previous research has demonstrated that a moderate flow of pedestrians can offer joggers increased opportunities for social interaction, thereby enhancing the social enjoyment derived from jogging [26]. Thus, controlling the flow of people on the greenway can enhance the vitality of the environment without causing a sense of crowding. In routine management, surveillance cameras and crowd risk assessment are suggested to be employed to monitor and guide the flow of people.

5.2.3. GVI and NVI

Previous studies have suggested that natural elements are very important for the perceived attractiveness of jogging environments; however, in this study, the impact of the GVI and NVI on PJS was lower than that of the artificial elements and spatial factors, possibly because the sample greenways had a relatively low variation in GVI [67].

However, the nonlinear curve shows that joggers prefer to jog in environments with a greater abundance of greenery. This is consistent with the research by Boakye and Amram, who found that green environments increase the frequency, attractiveness, and satisfaction of jogging [68]. When the GVI is less than 0.5, the scarcity of plants in the greenway leads to a lack of aesthetic appeal, which in turn inhibits PJS. When the GVI exceeds 0.85, an excessive amount of greenery can create an eerie atmosphere, and a high GVI can lead to a sense of insecurity among users [69]. When the GVI is between 0.5 and 0.85, an appropriate amount of greenery provides natural exposure for joggers, attracting them to engage in jogging, and thus also promotes PJS. For greenway design, the spatial layout of vegetation along the route should be to avoid over-concentration or dispersion. Additionally, the GVI can be altered by varying the layers of trees, shrubs, and ground cover plants. Regular trimming and maintenance of the vegetation should be conducted to maintain its good growth and appearance in order to keep the GVI within an appropriate range.

In this study, the NVI includes both vegetation and water, so the nonlinear correlation curve of the NVI should be consistent with that of GVI. However, the nonlinear correlation curve of the NVI is more gently influenced by the curve of water, and due to the lower frequency of water occurrence, its relative importance is not high.

5.2.4. Surface Material and Stairs

The surface material and stairs both relate to the path itself. Greenways with surfaces suitable for jogging and without stair-like obstacles promote PJS [25,32]. For surface materials, plastic or asphalt can prevent damage to joggers’ knees during exercise, facilitating jogging. In contrast, materials such as concrete and brick are not only prone to damage but also may impact joggers’ joints due to their high hardness. Previous studies have also demonstrated that the surface material significantly influences jogging, with asphalt pavement paths playing a pivotal role in augmenting the intensity and allure of such activities [25]. When selecting surface materials, appropriate materials like plastic or asphalt should be used to pave the surface, avoiding injury to joggers and thus affecting the jogging experience.

The presence of stairs on a jogging path can interrupt the continuous process of jogging, disrupt the jogger’s rhythm, and sometimes even cause injury to the jogger [66], thereby negatively impacting the jogging experience. In the design of greenways, the appearance of stairs should be avoided. When encountering changes in elevation, the difference can be managed by grading the slope, which not only makes the jogging process more continuous but also increases the variability of the environment and enhances interest in jogging.

5.2.5. EI and Sky

The EI and Sky both relate to joggers’ perception of openness. For the EI, a value below 0.1 provides shade for joggers, promoting PJS, while a value exceeding 0.1 can give joggers a sense of oppression, thereby inhibiting PJS. This is consistent with previous research, which found that there is an optimal design value for the degree of enclosure in enhancing road jogging intensity [23]. For greenway design, the EI could be controlled by reducing the use of tall fences and walls and using plants or perforated landscape walls for separation, thereby improving the jogging experience.

The visual ratio of sky also represents the degree of openness of the environment. When it is over 0.08, a larger area of sky exposes joggers to sunlight, thereby hindering thermal comfort and environmental satisfaction. When it is less than 0.08, a smaller area of sky prevents the space from being completely enclosed, providing joggers with light and security while also increasing their natural experience. Previous studies have shown that a certain degree of sky openness may promote jogging [23], which is consistent with our research findings. In greenways with high PJS, tall trees can be planted to naturally form shaded and rain-proof areas along the route, reducing direct sunlight with a high visual ratio of sky. This not only increases green coverage but also enhances the ecological value and aesthetic effect of the greenway.

5.3. Limitations and Prospects

Firstly, this study solely captured daytime greenway videos during sample collection, resulting in a diminished safety contribution value. As some joggers exclusively engage in jogging at night, the perception of nighttime joggers should not be disregarded. Future research should comprehensively consider data on the nocturnal greenway environment during sample collection, choosing a semantic segmentation model with higher accuracy to accurately identify nighttime environmental characteristics. Secondly, regarding the data collection for scoring the perception dimension, it should be noted that the subjective evaluation data in this study are solely based on judgments from college-student jogging enthusiasts, which may not fully represent the general public’s perceptual judgment. Therefore, future research could enhance reliability by increasing the number of raters and diversifying their identities to ensure a more comprehensive and accurate assessment of perceptual judgment. Thirdly, this study unveiled the significance of dynamic characteristics at the perceptual level; however, it lacked methodologies for quantifying dynamic attributes at the physical level. Future investigations should prioritize the development of computational approaches to assess dynamic characteristics. Furthermore, due to inherent research constraints, this study has not yet undertaken a longitudinal investigation into the pre- and post-perceived jogging supportiveness (PJS) of representative greenways. Therefore, future research should consider conducting a longitudinal study on a specific case to establish empirical evidence supporting the practicality of the study.

6. Conclusions

Utilizing a multi-method approach, this study identified the key characteristics that influence the perceived jogging supportiveness of greenways from both perceptual and physical perspectives. Furthermore, it revealed the nonlinear effects of these characteristics and proposed specific design suggestions for urban greenways that promote jogging.

On the perceptual level, the research findings revealed the following:

(1)

Higher levels of continuity, naturalness, and vitality were found to be associated with greater perceived jogging supportiveness (PJS). To enhance the PJS, designers should pay attention to the continuity of the greenway path, reduce interference, design a more natural greenway landscape, and increase the vitality of the greenway by enriching the landscape color.

(2)

Facility affordance, scale, culture, openness, and brightness exhibited optimal ranges, thus necessitating a careful control within their respective design parameters.

At the physical level, the research findings revealed the following:

(1)

An excessive number of people and vehicles were observed to interfere with jogging activities; conversely, incorporating more natural elements and utilizing comfortable path materials can enhance the experience of jogging on greenways. Therefore, it is necessary to design human–vehicle separation, enrich natural elements, and choose materials such as plastic and asphalt to lay roads.

(2)

Additionally, an overabundance of artificial facilities may lead to discomfort during jogging. Therefore, enclosures such as fences and walls are reduced in the design.

(3)

Moreover, it is crucial for a jogger-oriented greenway design to prioritize meeting behavioral needs rather than affective or aesthetic needs.

By collecting 230 video samples of greenways in Fuzhou City and analyzing them using the XGBoost framework and SHAP values, this empirical research provides practical reference for understanding the relationship between environmental characteristics and perceived jogging supportiveness (PJS). This study proposes design suggestions based on the impact of environmental characteristics on PJS, which can help to design more effective greenways that promote jogging. This research not only presents a theoretical framework but also supports it with empirical data, providing a scientific basis for urban planning and greenway design. Through optimizing urban greenways, this study helps increase their utilization rate, thereby positively impacting public health and aligning with the global trend of promoting healthy lifestyles.