Pre-processing urban landscape structure data
To construct the urban structure dataset, we used a 2008 1-meter
resolution land cover dataset (City of Philadelphia, 2011), The Property
Assessment dataset from the Philadelphia Office of Property Assessment
(City of Philadelphia, 2015b) indicating number of floors in buildings
for each tax lot in the city in tabular format, and the Philadelphia
Department of Water parcels dataset. We joined the property assessment
tabular data to the parcels dataset using unique parcel IDs and created
a 1-meter resolution raster dataset from the Number of Floors field in
the Property Assessment dataset. Number of floors were classified into
three categories: lowrise (1–3 stories), midrise (4–9 stories) and
highrise (>9 stories) (Larondelle et al., 2014; I. D.
Stewart & Oke, 2012). We then combined it with the land cover raster
dataset, by replacing all building land cover pixels with a value
representing building height category to create our basic urban
structure dataset.