A Continuous Infrastructure Index for Mapping Human Settlements

Christopher Small, Columbia University, cs184@columbia.edu (Presenter)
Son V. Nghiem, NASA JPL, son.v.nghiem@jpl.nasa.gov

The Landsat program provides more than three decades of decameter resolution multispectral observations of the growth and evolution of human settlements and development worldwide. These changes are often easy to observe visually but accurate repeatable quantification at Landsat’s resolution has proven elusive. In part, this is a consequence of the multi-scale heterogeneity and diversity of settlements worldwide. Mapping settlement extent is also confounded by the lack of a single, physically based, definition of what constitutes urban, peri-urban and other types of settlement. We attempt to resolve both of 
these challenges by characterizing built environments in terms of their distinctive physical properties. This can be accomplished
by combining multi-temporal optical reflectance with synthetic aperture radar backscatter measurements to identify combinations of physical properties that distinguish built environments from other types of land cover. Three well-known examples include 
an abundance of impervious surface, persistent deep shadow between buildings and high density of corner reflectors at meter 
to decameter scales. At optical wavelengths, spectral properties of land cover can be represented using standardized spectral endmember fractions to represent combinations of the most spectrally and functionally distinct components of land cover such as soil and impervious substrates, vegetation, water and shadow. The spectral similarity of soils and impervious substrates that makes thematic classifications error prone can be resolved by using multi-season composites of spectral endmembers to distinguish spectrally stable impervious substrates from temporally variable soil reflectance resulting from seasonal changes in moisture content (thus albedo) and fractional vegetation cover. By representing the diversity of anthropogenic land use as a continuous mosaic composition of land cover it 
is possible to quantify the wide variety of human settlements in a way that is physically consistent, repeatable and scalable. Our strategy is to develop and test algorithms to combine multi-season Landsat and Sentinel-2 optical multispectral imagery with SRTM and Sentinel-1 C-band radar backscatter imagery to produce a Continuous Infrastructure Index (CII) to identify and map changes 
in the extent of anthropogenic built environments (e.g. urban, suburban, exurban, peri-urban) worldwide between 2000 and 2015. Rather than attempting to map specific features or disparate components associated with built environments (e.g. impervious surfaces, buildings, roads),
 we characterize the combined optical and microwave response of a wide range of built environments to identify the physical properties associated with these features (e.g. spectral stability, persistent shadow, backscatter and its derivatives). We will then use the most persistent of these properties to derive an optimized index incorporating multiple characteristics measured by both optical and microwave sensors. Variations in relative density of stable substrate (impervious surface), building shadow and corner reflectors will be used to define a continuous space of built environment characteristics for different types of human settlement worldwide. Changes CII between 2000 and 2015 will track both vertical and horizontal growth as well as temporal evolution of settlement networks worldwide.

Presentation Type:  Poster

Session:  General Contributions   (Tue 4:35 PM)

Associated Project(s): 

• Nghiem, Son: Mega Urban Changes and Impacts in the Decade of the 2000s ...details

Poster Location ID: 120