Spatial analysis of the Surface Urban Heat Island

In this paper, we present a novel framework to characterize the complex spatial structure of the intra-urban heat island. Cities are known to be warmer than its surrounding areas because of the Urban Heat Island (UHI) phenomenon. However, due to the diverse and complex spatial geometries of cities themselves, the temperatures within vary widely. We take advantage of the well-established notion of fractal properties of cities, to characterize the complex structure of these hotspots. As a demonstrative case study, Land Surface Temperatures (LST) for Atlanta, GA, derived from Landsat 8 is used. From clustering analysis at multiple thermal thresholds, we show that the hotspots can be described as a case of percolating clusters. By comparing the area-perimeter fractal dimension at these thresholds, we find these clusters to be statistically self-similar. Furthermore, at the percolation threshold, the cluster size distribution is found to follow a power-law size distribution; and at a higher threshold, deviation from the power law is observed in the form of exponential tempering. We argue that the spatial distribution of the hotspots itself plays a significant role in the overall UHI and fractal analysis techniques lend themselves aptly to the characterization of the same. This has several further applications, such as targeted heat mitigation, assessment of health impacts, and energy load estimation.