The algorithmic and representational uncertainties in green space exposure measurement: conceptualizations and a suite of available methods for mobility-oriented studies

Multiple coupled green space data and representations from Earth observations are available for deriving green space exposure measurements, but they may not be equally effective in health outcome modeling. This indicates algorithmic and representational uncertainties in green space research. In this study, we conceptualized this methodological issue and systematically compared a suite of available green space representations for mobility-oriented green space exposure measurements. Five typical green space representations in Hong Kong were derived from either remote sensing or street view image (SVI) data. Multiple exposure measurements were then derived through a spatiotemporal accumulation approach using each green space representation. The results indicate that green space exposure measurements derived from different representations may have different magnitudes, inconsistent correlations between each other, contradictory associations with participants’ activity space sizes, and may or may not reveal the disparities in the characteristics of green space in different geographic contexts. Since different green space representations may be coupled with different health pathways, our study provided essential insights into properly conceptualizing green space representations to derive causally relevant and effective green space exposure measurements and to mitigate the algorithmic and representational uncertainties in a range of environmental health studies that consider green space an important factor in human health.