118 Thirty-eight urban regions
image processing in Arcview. The resulting image was printed in color with the
city in the approximate center. Preliminary determination of the boundaries
formany urban regions found that boundaries rarely extend in any direction
more than 100 km from center city, so images were printed with a diameter of
approximately 200 km.
After considerable consultation with GIS (geographic information systems)
specialists, I decided that hand measurements using planimeter and ruler on
clear plastic sheets over the satellite images was the optimal procedure. Suf-
ficient quality, accuracy, and consistency to accomplish the objectives seemed
unavailable by GIS analysis without considerably more time and resources than
were available. Especially worrisome was the uncertain ability to correctly dif-
ferentiate bycomputer the hundreds of land covers in urban regions world-
wide. Doubtless I would make some errors in image interpretation, but based
on 35 years of landscape ecology work and analyzing/ground-truthing maps and
GIS images in many nations, the decision to use manual rather than GIS mea-
surements ended up an easy one.
Withthe30mcell-size resolution on the satellite image, two-lane highways
were often invisible or hard to follow, whereas most of a multilane highway
length was clearly evident. Streams and small rivers disappeared in agricultural
and built landscapes, the predominant land covers in most urban regions, except
where wide stream-corridor vegetation was present. In forested landscapes, rivers
and streams less than about fifth-order (Wetzel2001,Kalff 2002) could seldom
be followed, because the open strip in the tree canopy was too narrow. The
rare forested wetland generally was not differentiable from forest. Hedgerows,
individual houses, and narrow two-lane roads normally were invisible.
In contrast, major airports, shipping/ferry ports, and large mines or quarries
were quite prominent. Normally I could differentiate: marsh from other open
areas; mangrove swamp from other woodland/forest; dammed reservoirs from
lakes; mountain ridges (but not hills); villages and hamlets; and direction of
surface-water drainage and stream/river flow.
Urban-region boundaries, metropolitan areas, and numerous specific objects
were marked on the large images (c. 70 ×100 cm). Spatial measurements (area,
shape, distance, and number) were then made for both marked attributes and
unmarked patterns. When new information and corrections became infrequent
foranurban region, the markings were copied with black pen onto a small
image (43×43 cm) of the region.
An illustrator then converted the major land covers marked into brightly
colored maps and computer-reduced the images for printing in this book. The
colors represent a balance between standard urban-planning practices and the
ability to differentiate land covers in a black-and-white photocopy of the image.