GPS-based street-view greenspace exposure and wearable assessed physical activity in a prospective cohort of US women

FAQ: Greenspace Exposure and Physical Activity

What was the primary objective of this study regarding greenspace and physical activity?

The primary objective of this study was to examine the associations between momentary, GPS-based street-view greenspace exposure and objective, momentary physical activity (PA) data. Unlike previous research that often used satellite-derived measures around residential areas or aggregated different greenspace types, this study aimed to capture ground-level views and day-to-day dynamic exposures to specific greenspace components (trees, grass, and "other greenspace" like flowers/plants/fields) and their immediate impact on PA. The goal was to provide more nuanced insights for urban planning and public health interventions.

What specific types of greenspace were analyzed, and what were the main findings regarding their association with physical activity?

The study analyzed three specific types of street-view greenspace: trees, grass, and "other greenspace" (plants, flowers, and fields).

• Trees: Momentary exposure to street-view trees was found to be inversely associated with physical activity, meaning an increase in visible trees was linked to a decrease in steps-per-minute. This small negative association suggests that trees, in the contexts studied, may not consistently promote PA.
• Grass: Momentary exposure to street-view grass showed a small, non-linear inverse association with physical activity. Specifically, increased grass exposure was associated with lower PA up to a certain percentile (around 10%), after which the association flattened. This implies that large grassy areas might support more sedentary activities rather than active recreation.
• Other greenspace (plants, flowers, fields): Conversely, momentary exposure to "other greenspace" was positively associated with physical activity, meaning an increase in these elements was linked to an increase in steps-per-minute. This category might indicate well-maintained, aesthetically pleasing environments that encourage higher PA levels.

What were the key methodological advancements of this study compared to previous research on greenspace and PA?

This study introduced several key methodological advancements:

• GPS-based momentary exposure: Instead of relying on greenspace measures around a static residential address, the study linked 10-minute GPS observations from smartphones to concurrent street-view imagery, capturing dynamic, ground-level exposure as participants moved throughout their day. This addresses the "uncertain geographic context problem" where exposure beyond residential areas is often overlooked.
• Street-view imagery and deep learning: It leveraged street-view imagery combined with deep-learning algorithms (Pyramid Scene Parsing Network) to segment images at a pixel level into specific categories (trees, grass, plants, flowers, fields). This allowed for the differentiation of greenspace components, which was a significant improvement over aggregate measures like Normalized Difference Vegetation Indices (NDVI) used in many previous studies.
• Objective PA measurement: Physical activity was objectively measured using Fitbit™ wearable devices, collecting minute-level step counts, which reduced the likelihood of recall bias associated with self-reported measures.
• Intensive longitudinal design: The study collected data from participants over four 7-day periods throughout a year, allowing for the examination of seasonal and time-variant effects and enabling intra-individual comparisons.

Were there any modifying factors that influenced the observed associations between greenspace and physical activity?

Yes, the study found that the associations between street-view grass exposure and physical activity were modified by season, geographic region, and neighborhood socioeconomic status (SES):

• Season: The inverse association between grass exposure and PA was strongest during spring. The researchers speculate this might be due to a higher proportion of social and sedentary activities (e.g., picnics) on grassy lawns in spring compared to more active recreation in summer or fall.
• Region: The inverse association with grass was most pronounced in the Northeast region of the US.
• Neighborhood SES: The inverse association between grass exposure and PA was only observed in neighborhoods with higher SES (3rd and 4th quartiles), with no statistically significant association in lower-SES neighborhoods. Higher SES neighborhoods, often having better aesthetics and more public spaces, might facilitate sedentary recreational activities in grassy areas.

No significant seasonal, regional, or SES differences were observed for the associations of street-view trees or "other greenspace" with PA.

How did the study account for potential biases and validate its findings?

The study conducted several sensitivity analyses to test the robustness of its findings and account for potential biases:

• Selective daily mobility bias: To mitigate this bias (where more active individuals might selectively visit environments conducive to exercise), GPS data was restricted to locations within participants' usual activity spaces. The core findings remained consistent.
• Excluding work hours: Associations were examined during non-work hours by excluding GPS data near geocoded workplace addresses.
• Active transportation/recreation: Data was limited to GPS points with velocities corresponding to walking and running, reducing the inclusion of sedentary activities or driving time.
• Restricting cohort: Analyses were repeated on a more stringent cohort subset with more complete GPS data, confirming robustness.
• Adjusting for other metrics: The main models were also re-run adjusting for satellite-based greenness (NDVI) and street-view sidewalks, and using larger buffer sizes (500-meter and 1,000-meter), demonstrating consistency.

These sensitivity analyses largely supported the main findings, indicating their robustness.

What are the practical implications of these findings for urban planning and public health initiatives?

The findings have important implications for urban planning and public health initiatives focused on promoting physical activity:

• Prioritize "other greenspace": The positive association with "other greenspace" (flowers, plants, landscaped areas) suggests that investing in and maintaining these features can enhance the perceived quality and aesthetic appeal of greenspaces, thereby encouraging higher PA levels.
• Contextualize trees and grass: The inverse associations with trees and grass indicate that these elements alone may not consistently promote active PA. Urban planners should consider the surrounding infrastructure (e.g., presence of sidewalks, trails) and design considerations to ensure these areas facilitate, rather than hinder, active movement. Large grassy areas might be more conducive to sedentary activities, so their design should be considered in light of desired activity outcomes.
• Nuanced interventions: The study highlights the need for a nuanced approach to greenspace design, moving beyond aggregate "greenness" measures. Different types of greenspace serve different purposes and influence behavior differently.
• Tailored strategies: The observed effect modifications by season, region, and SES for grass exposure suggest that greenspace interventions might need to be tailored to specific contexts and populations to be most effective.

What are the main limitations of this study, and what are the recommendations for future research?

Limitations:

• PA measurement scope: Step count as a proxy for PA does not capture all forms of physical activity (e.g., weightlifting, cycling, gardening, swimming).
• Device wear time: Participants did not wear devices at all times, potentially missing some activity data.
• Greenspace temporal variation: Street-view metrics were derived from annual snapshots, potentially missing finer seasonal variations in greenspace appearance.
• Street-view perspective: Images only capture greenspace visible from the road, potentially missing parks and other greenspaces away from roads or accounting for slopes.
• Cohort generalizability: The Nurses' Health Study 3 cohort primarily consisted of upper-middle-class White women, limiting the generalizability of findings to more diverse populations (e.g., lower-income, different racial/ethnic groups).
• Selective daily mobility bias: While addressed, fully distinguishing incidental exposure from self-selected exposure for PA remains a challenge without detailed travel diaries.

Recommendations for Future Research:

• Confirm in diverse populations: Future research should confirm these results in cohorts with greater racial, ethnic, and socioeconomic diversity.
• Explore mechanisms: Further studies should explore the specific mechanisms through which different greenspace components influence physical activity behavior.
• Comprehensive PA data: Incorporate data on other forms of PA beyond step counts where feasible.
• Travel diaries/mobility surveys: Implement travel diaries or mobility surveys to better understand trip purposes and address selective daily mobility bias more comprehensively.
• Detailed greenspace quality: Investigate the specific design and maintenance features of "other greenspace" that contribute to its positive association with PA.

How significant were the observed associations in terms of daily steps?

Although some associations were described as "small" or "relatively small," their cumulative impact over a day could be meaningful:

• Grass: An IQR increase in street-view grass exposure (10.6%) was associated with a decrease of 0.59 steps-per-minute. For an individual active for 16 waking hours a day, this translates to approximately 568 fewer steps per day.
• Other greenspace: Conversely, an IQR increase in "other greenspace" (1.2%) was associated with a 1.99 steps-per-minute increase. This could result in a daily increase of approximately 1,910 steps.

The study notes that an increase of 1,910 steps per day is a level associated with notable health benefits, including reduced cardiovascular risk. This emphasizes that even seemingly small per-minute associations can lead to significant daily step changes and have important public health implications.