Typological distinction of remotely sensed metrics of neighborhood vegetation for environmental health intervention design

Abstract

AbstractThe extent to which urban vegetation improves environmental quality and affects the health of nearby residents is dependent on typological attributes of “greenness”, such as canopy area to alleviate urban heat, grass to facilitate exercise and social interaction, leaf area to disperse and capture air pollution, and biomass to absorb noise pollution. The spatial proximity of these typologies to individuals further modifies the extent to which they impart benefits and influence health. However, most evaluations of associations between greenness and health utilize a single metric of greenness and few measures of proximity, which may disproportionately represent the effect of a subset of mediators on health outcomes.To develop an approach to address this potentially substantial limitation of future studies evaluating associations between greenness and health, we measured and evaluated distinct attributes, correlations, and spatial dependency of 13 different metrics of greenness in a residential study area of Louisville, Kentucky, representative of many urban residential areas across the Eastern United States. We calculated NDVI, other satellite spectral indices, LIDAR derived leaf area index and canopy volume, streetview imagery derived semantic view indices, distance to parks, and graph-theory based ecosystem connectivity metrics. We utilized correlation analysis and principal component analysis across spatial scales to identify distinct groupings and typologies of greenness metrics.Our analysis of correlation matrices and principal component analysis identified distinct groupings of metrics representing both physical correlates of greenness (trees, grass, their combinations and derivatives) and also perspectives on those features (streetview, aerial, and connectivity / distance). Our assessment of typological greenness categories contributes perspective important to understanding strengths and limitations of metrics evaluated by past work correlating greenness to health. Given our finding of inconsistent correlations between many metrics and scales, it is likely that many past investigations are missing important context and may underrepresent the extent to which greenness may influence health. Future epidemiological investigations may benefit from these findings to inform selection of appropriate greenness metrics and spatial scales that best represent the cumulative influence of the hypothesized effects of mediators and moderators. However, future work is needed to evaluate the effect of each of these metrics on health outcomes and mediators therein to better inform the understanding of metrics and differential influences on environments and health.