Measuring green space exposure for epidemiological studies: moving from 2D to 3D indicators

The presence of green spaces has been associated with improved physical and mental health and wellbeing. The available evidence on the health effects of green spaces have relied on a range of methods to estimate the exposure to green spaces, including questionnaires, land use maps or satellite derived vegetation indices such as NDVI (Normalized Difference Vegetation Index). Although they are acceptable proxies of such exposure, they rely solely on 2-dimensional (2D) data and thus can have considerable inaccuracies when characterizing diverse set of vegetation types in combination with grey (built-up) spaces, which is typical of urban environments. For example, these 2D indicators imprecisely assess tree volumes (or biomass) and therefore hardly identify their direct and indirect positive effects (e. g. carbon sequestration or temperature reduction). Moreover, these 2D indicators are not able to provide a holistic view of the combination of green and grey spaces in urban settings. To overcome this gap, we developed a set of novel three-dimensional (3D) indicators derived from airborne Light Detection and Ranging (LiDAR) acquired in 2008 and 2010 over the metropolitan area of Rome (Italy). In particular, we extracted volume of green features e.g. shrubs and trees (Green volume [m3/ha]), volume of buildings (Gray volume[m3/ha]), a newly developed index called Normalized Difference Green/Gray Volume index (NDGG) as well as indicators of the tree count. We compared the 3D indicators with two widely used 2D indicators for characterizing green and grey spaces (i.e., NDVI and Imperviousness) in different buffers around 79140 address points in the city of Rome (Italy). For green indicators, we found that the correlations between NDVI and Green Volume were 0.47 (50m buffer) and 0.33 (300m buffer) while the correlations between NDVI and number of trees were 0.56 (50m buffer) and 0.58 (300m buffer). For gray indicators, the correlations between Imperviousness and gray volume were 0.62 (50m buffer) and 0.79 (300m buffer). For NDGG, the correlations were higher with both NDVI (0.76 and 0.83 for 50m and 300m buffers) and Imperviousness (-0.75 and -0.83 for 50m and 300m buffers). Our results showed that compared to 2D indicators, the 3D indicators can have potential advantages in characterizing green and grey spaces, especially in regard to green features which can be highly heterogeneous, particularly in complex urban landscapes such as the city of Rome.