Our study sought to build upon studies examining aspects of the natural environment that influence childhood obesity risk [13–16]. Consistent with previous studies [14–16], we found that exposure to higher levels of neighborhood greenness may reduce the risk for childhood obesity due to increased physical activity. Specifically, after controlling for socio-demographic and parental support factors, children residing in neighborhoods with higher levels of neighborhood greenness were more likely to engage in outdoor physical activity. To further explore this relationship, we examined whether physical activity behavior (i.e., total outdoor playing time, active vs. quiet play) was influenced by neighborhood greenness. Children living in neighborhoods with the highest levels of neighborhood greenness were more likely to engage in outdoor physical activity (multi-players), and engaged in similar bouts of active and quiet outdoor play. This suggests that exposure to greener neighborhoods encourages children to spend more time outdoors where they may reap both physiological and cognitive benefits [13, 34]. Conversely, children with the lowest levels of neighborhood greenness were least likely to spend time playing outdoors, engaging in active or quiet play (rainy day kids). While neighborhood greenness influenced levels of physical activity among these preschool-aged children, as Cleland et al  observed, parental support factors such as engaging in physical activity with children also plays a role. This may explain why sporty kids were more likely to engage in much more active outdoor physical activity compared to rainy day kids, although both groups lived in areas with similar levels of neighborhood greenness (Table 2). Moreover, although sandbox lovers had higher levels of neighborhood greenness than sporty kids, their parents were least likely to transport them to opportunities for physical activity (Table 2). It is important to note, however, that perhaps the personality of some of the children, such as the rainy day kids, with the tendency to engage in limited outdoor activity, may also influence parents to spend less time transporting them to opportunities for physical activity.
Since the research literature suggests that boys are typically more active than girls between pre-school age and adolescence, we were not surprised to find that boys tended to engage in more physical activity, compared to girls in our sample (Tables 1 and 2) [31, 36, 37]. Socialization by gender, where boys may be encouraged to engage in more physical activity may explain this difference.
While our findings support previous studies exploring the influence of neighborhood greenness on childhood obesity risk, there are some limitations. Our study was cross-sectional, and focused only on levels of physical activity, rather than BMI. Moreover, our measure of physical activity relied on parental reports, rather than objectively measured data collected from accelerometers. Notwithstanding, the quantification of physical activity among preschoolers using accelerometers has been wrought with challenges due to the inability to account for differences in motor and cognitive development that may influence movement patterns [38–40]. While accelerometers seem to be far more effective measuring physical activity among school-age children, standardized methods are still being developed for preschoolers .
Our sample size may have also been too small to give us more robust estimates, as indicated by the wide confidence intervals for some of the results of our regression analyses. Our omission of certain measures of the neighborhood environment may have also influenced our estimates, such as neighborhood walkability, which has been shown to influence overweight among pre-school aged children . We also did not control for neighborhood measures of socioeconomic status, which have been shown to influence childhood overweight, independent of neighborhood greenness . When we included measures of neighborhood SES, however, they were highly correlated with NDVI, so we did not include them in our models to avoid multi-collinearity. Interestingly, when we ran our models using only measures of neighborhood SES to determine whether the NDVI was strictly serving as a proxy for SES, we did not find the same relationship with levels of physical activity. Specifically, while higher levels of neighborhood greenness were associated with higher levels of physical activity, the same did not hold true for higher levels of neighborhood SES. This observation may be due to the fact that the study area encompasses micro-urban and semi-rural areas, where neighborhood SES may be less important for access to greenspace, compared to many urban environments in the US. We also did not account for seasonality based on the time of the year that parents completed the survey. A study by Finn et al. , however, found that seasonality did not influence the total physical activity time of pre-school aged children. In addition, our own analysis (not shown) did not indicate that seasonality influenced levels of physical activity reported by parents.
Another limitation is the use of the area around the home address as the spatial unit for measuring neighborhood greenness. However, we believe that using the home address of pre-school aged children in this sample was the best spatial proxy for measuring the influence of neighborhood greenness on physical activity for several reasons. First, there is some evidence  to suggest that children usually play close to where they live. As such, we attempted to account for areas close to where children live that would encompass nearby playspaces in addition to their backyard, for example. Of course, it is still possible that parents could take their children to places outside of their immediate neighborhoods to engage in active play. Given what the authors know of the study area, however, many families of young children do in fact regularly utilize their neighborhood playspaces. Second, approximately 20% of children in our sample did not attend day care every day, and only 30% of parents reported having their children in day care 40 hours a week. As such, we could only assume that the children were also spending quite a bit of time at their home address. Third, parents were asked to separately report on weekend and weekday physical activity, thus the home address is the only spatial unit that is consistent across all 7 days of the week. Given what we knew of day care attendance in our sample, and park/playspace utilization in the study area, we believe that we used the best spatial proxy for measuring neighborhood greenness based on our study design and resources. We do concede, however, that the use of GPS to actually track the children during the day, or using an activity space log would have more accurately captured the play spaces of the children in our sample.
Another issue that we also explored was the potential cluster effect of day care centers. However, our preliminary analyses showed an intraclass correlation of less than .01. As such we felt confident that the relationships observed in our study were not overestimated.
Despite these limitations, our study also has several strengths. We used an objective measure of neighborhood greenness, and examined both total outdoor playing time and physical activity patterns among children. Our study is also one of the first to examine the influence of the natural environment on childhood obesity risk that focuses exclusively on pre-school aged children and accounts for physical activity behaviors, socio-demographic and parental support factors. By focusing exclusively on pre-school aged children, this allows us to tailor efforts on a critical developmental stage for targeting interventions.