Multisensor characterization of urban morphology and network structure

Abstract

The combination of decameter resolution Sentinel 2 and hectometer resolution VIIRS offers the potential to quantify urban morphology at scales spanning the range from individual objects to global scale settlement networks. Multi-season spectral characteristics of built environments provide an independent complement to night light brightness compared for 12 urban systems. High fractions of spectrally stable impervious surface combined with persistent deep shadow between buildings are compared to road network density and outdoor lighting inferred from night light. These comparisons show better spatial agreement and more detailed representation of a wide range of built environments than possible using Landsat and DMSP-OLS. However, they also show that no single low luminance brightness threshold provides optimal spatial correlation to built extent derived from Sentinel in different urban systems. A 4-threshold comparison of 6 regional night light networks shows consistent spatial scaling, spanning 3 to 5 orders of magnitude in size and number with rank-size slopes consistently near -1. This scaling suggests a dynamic balance among the processes of nucleation, growth and interconnection. Rank-shape distributions based on pArea/Perimeter of network components scale similarly to rank-size distributions at higher brightness thresholds, but show both progressive then abrupt increases in fractal dimension of the largest, most interconnected network components at lower thresholds.