Investigating the implications of using alternative GIS-based techniques to measure accessibility to green space

A large body of research has examined relationships between accessibility to green space and a variety of health outcomes with many researchers finding benefits in terms of levels of physical activity and relationships with levels of obesity, mental health, and other health conditions. Such studies often use spatial analytical techniques to examine relationships between distance to such spaces and health data collated at an individual survey respondent’s home address or, more commonly, derived from area-based census measures summarised at a centroid. Generally, such measures are becoming more sophisticated and have moved on from the use of straightforward Euclidean-based measures to those based on network distance. However, few studies tend to use a combination of approaches or seek to establish the implications of incorporating alternative measures of accessibility on potential relationships. Using a database of green spaces (and associated attributes) and a detailed network dataset for the city of Cardiff, Wales, we examine the sensitivity of findings to the ways in which different metrics are calculated. This is illustrated by examining the variations in association between such metrics and a census-based deprivation index widely used in health studies to measure socioeconomic conditions. Our findings demonstrate that not only will the distances to green spaces vary according to the methodologies adopted but that any study that aims to investigate relationships with attributes of the <i>nearest </i>green space should acknowledge that matches may vary widely according to the techniques used. We conclude by warning against the use of inappropriate methodologies in examining access to green space which may directly influence directions (and levels) of association and hence may limit their relevance in wider geographical contexts. <br> <br><b>Keywords:</b> green space, accessibility, GIS, spatial data, representation, network distances