Evaluating street view exposure measures of visible green space for health research

Urban green space, or natural environments, are associated with multiple physical and mental health outcomes. Several proposed pathways of action for these benefits (e.g., stress reduction and attention restoration) require visual perception of green space; however, existing green space exposure measures commonly used in epidemiological studies do not capture street-scale exposures. We downloaded 254 Google Street View (GSV) panorama images from Portland, Oregon and calculated percent of green in each image, called Green View Index (GVI). For these locations we also calculated satellite-based normalized difference vegetation index (NDVI), % tree cover, % green space, % street tree buffering, distance to parks, and several neighborhood socio-economic variables. Correlations between the GVI and other green space measures were low (-0.02 to 0.50), suggesting GSV-based measures captured unique information about green space exposures. We further developed a GVI: NDVI ratio, which was associated with the amount of vertical green space in an image. The GVI and GVI: NDVI ratio were weakly related to neighborhood socioeconomic status and are therefore less susceptible to confounding in health studies compared to other green space measures. GSV measures captured unique characteristics of the green space environment and offer a new approach to examine green space and health associations in epidemiological research.