Abstract

Environmental inequalities are a common characteristic of urban areas. Environmental inequality is the unequal spatial distribution of environmental risks and goods among social groups. As environmental inequalities are inherently a spatial matter the choice of scale is essential for correctly understanding inequality issues and for designing proper and effective mitigation policies. However, the potential effects of scale of analysis on inequalities results have largely been underestimated in the assessment of environmental inequalities, leading to contradictory results from different studies. In this study we assess the patterns of environmental inequalities and associated scale issues in the city of Santiago (Chile) using a hierarchical multiscale approach. Our approach focuses on the analysis of spatial relationships between three environmental (i.e., surface temperature, air pollution, vegetation cover) and two socio-demographic variables (i.e., household wealth, population density) on multiple grain sizes and extents. We used census data, remote sensing data, and air pollution monitoring stations to generate raster layers at five grain sizes and five nested extents. We tested for inequalities through Pearson correlation analysis resulting in a total of 1530 assessed relationships. Our results show that environmental inequalities are a prevalent phenomenon in the city of Santiago, but the details of these inequalities are highly scale dependent. Changing the grain size and extent of analysis do not only affect the strength of relationships between socio-demographic and environmental variables, but also the spatial distribution of environmental inequalities across the urban landscape. Therefore, due to the scale dependence of assessment results, researchers and decision-makers should be extremely careful when interpreting their findings and translating them into policy making. If the scale dependency of environmental inequalities is not taken into account, policy interventions may be largely ineffective because the scale at which interventions are designed may not match the scale at which inequalities are generated. (C) 2016 Elsevier Ltd. All rights reserved.