Selecting cities, determining boundaries, mapping regions 117frequently helped delineate a region boundary. However, boundary location
seemed dependent on the relative size of the core city vs. the other city. If the
other city was larger, a point 40 % of the distance from core city to other city
wasarbitrarily marked as a preliminary estimate of boundary location. If the
other city was smaller and >250 000 population, a 60 % point was marked. If the
outer city was smaller and also <250 000 population, the 70 % point was marked.
One-day recreation or tourism sites were also significant boundary delineators,
theidea being that both the core city and the sites are important economically
and culturally to each other, and cannot be too far apart relative to the trans-
portation system. Major biodiversity areas, either impacted by or dependent on
protection by people of the city, sometimes helped delineate boundaries. Out-
lines of the drainage basin around major water supplies were often important
in determining urban-region boundaries.
Finally, where none of the preceding seemed important, a radius of approx-
imately 100 km was chosen. Partly this was because almost all the other delin-
eated urban-region boundaries were in the 70--100 km range. And partly it was
selected as a typical maximum distance on a paved highway with traffic that
large numbers of people would travel back and forth in one day for shopping,
recreation, and so forth. Even around remote cities with unpaved radial roads,
10 0kmmaybe a reasonable distance, for example, for once-a-week shopping or
business trips.
Food in urban regions
While literature, consultations, and direct observations helped under-
stand urban regions, the core analysis were measurements of spatial patterns
on large remotely sensed images (c. 70 × 10 0cmand 1:200 000 scale). Consistent
base maps for urban regions using Landsat satellite geospatial data, with a 30 m
cell or pixel size, were generated from the Earth Science Data Interface (2006
website) of the University of Maryland’s Global Land Cover Facility, College Park,
Maryland, USA. The data were organized in color spectra that provided the abil-
ity to separate red, green, and blue, as well as more advanced bands that could
separate urban areas from forest, meadow areas, and the like. The data were
finally manipulated in the ESRI ArcGIS 9.1 geographic information system. The
Arc Tool ‘‘Composite Bands” allowed for an image to be created, and point scal-
ing was used to create a number of different images. All flights for the remotely
sensed images of the 38 urban regions were in 2001±1year (except Santiago,
Tegucigalpa, and Sapporo, flown in late 1999).
The mapping process was a series of steps that created an aerial image as part
of a comprehensive database (see AppendixI). The steps began by extracting
therawband of color from the Earth Science Data Interface and ended with