883 
just 1972. Left, 
Center is in the 
jusing and then 
le floodplain of 
anquilla, 
ional Airport 
ti resolution 
s suburb, 
highway skirts 
Magdalena River 
5 port 
, are visible 
ds obscure some 
ty, but a great 
.it the urban 
Figure 5. Digitally processed Landsat TM subscene collected ovpr San Francisco on May 2, 1984. upper left Band 
3, visible red, 1630 - 690 nm; upper right Band 4, near reflective infrared, /bU-900 nm; lower left, Band /, 
short-wave infrared, 2080-2350 nm; lower right; Band 6, thermal infrared; 10,400-12,500 nm. Streets and land 
cover patterns are apparent in the 30 m resolution imagery, as are golf course,fairways, piers, and large 
industrial buildings. 
water, etc.-- wmch results in different reflectance 
values (Figure 3). In a computer classification of 
Landsat MSS data collected over the Seattle - Tacoma 
area (Gaydos & Newland, 1978), land cover classes 
included Commercial-Industrial, Residential, 
Pasture-Grass, Cropland, several types of forested 
land, Wetland, Barren Land, and Quarries- 
Transitional. Analysis of Washington, D.C. Landsat 
data yielded similar urban classes, including 
Commercial-Industrial-Services, Paved Surfaces, 
Older Residental, Newer Residential, Disturbed Land, 
Improved Open Space, Agriculture, and Forested 
Land/Brushi and (Gaydos & Wray, 1978). 
The urban-rural fringe presents a problem to the 
80 m MSS data, for two reasons. First, residential 
and other land use developments are in a transition 
al state. From land clearing through construction 
through landscaping tnrough maturing of vegetation 
and graying of concrete, a new development has a 
changing appearance both spectrally and spatially. 
Secondly, the rural land use adjacent to the 
urban-rural fringe is in either of three states: 
natural vegetation (forested land, shrub, 
grassland), agriculture/truck farming, or vacant 
land. Natural vegetation is the most predictable, 
spectrally, of the three states, and usually 
provides the best contrast with urban land uses. 
The agricultural land, unfortunately, undergoes 
significant signature changes as the land changes 
from plowed to emerging crop to mature crop to 
harvested crop to fallow land. In the Great Lakes 
region of North America, agricultural land contrasts 
well with urban areas when the crops are at the 
height of the growing season. The third state, 
vacant land, may not contrast well with an adjacent 
urban development. Forster (1980) provides detail'^ 
insight on residential land cover, and Jensen & Toll 
(1982) suggest that texture measures may be helpful 
in mapping the urban-rural fringe zone. 
While transition at the urban-rural fringe causes 
difficulty for land cover computer-assisted 
classification and mapping, scraping the earth for 
new development is detectable under most data 
collection circumstances. An application of Landsat 
to urban land cover change detection (using Landsat