Image Processing Reference
census is an enormously expensive and time-consuming undertaking, perhaps in the
future satellite imagery might prove a dependable method for inter-censal estimates
of national populations and urban growth.
Night-time lights imagery have also been useful as a data source for generating
urban indicators that may enable better understanding of urban populations than
demographic data alone. Lo ( 2002 ) used zones of radiance extracted from OLS data
and a Triangulated Irregular Network (TIN). Surface area and volume were also
generated and input into an allometric growth model used to estimate urban indica-
tors for 35 cities in China. This method proved to be very accurate when used to
estimate total non-agricultural population, a commonly used demographic measure
Gallo et al. ( 2004 ) also used DMSP OLS data of the United States to study
changes in levels of night-time light emitted between 1992/1993 and the year 2000.
In concert with the night-time lights data, the researchers also computed a time-
integrated Normalized Difference Vegetation Index (TINDVI) for the same period
across time and space. Those places where increased levels of emitted night-lights
occurred also showed a decrease in the TINDVI. The authors suggest that areas of
land cover change and increased night-light brightness have likely undergone
DMSP OLS data have also been used to map human settlements (Tobler 1969 ;
Elvidge et al. 1997 ) and delineate urban centers (Imhoff et al. 1997 ; Henderson
et al. 2003 ). In an effort to create a scale-adjusted index of urban sprawl, Sutton
( 2003 ) used DMSP OLS radiance calibrated night-time lights imagery of the
United States as a proxy for the extent of urban areas. Urban population counts
from the 1990 census were then combined with the information on urban area in a
regression model to allow for city-to-city comparisons of urban sprawl.
There have been some technical challenges to using the DMSP OLS data.
Concern about the overestimation of urban extent and shift in locational accuracy
is due in part to the difficulty in selecting light thresholds. One approach has been
to use an alternative higher resolution image, i.e., Landsat TM, as a calibration
data set (Henderson et al. 2003 ). The authors found that cities at different levels
of development require different light (both stable and radiance) thresholds in
order to assess urban size and location most accurately. Another issue has been
that of pixel saturation due to the 6-bit radiometric resolution of the DMSP OLS
data (Sutton 1997 ). Finally, an obvious limitation is the low spatial resolution
of the imagery. The use of DMSP OLS data has been successful at the global,
regional and city scale; however their effectiveness has not been shown at any
Social Science Applications
Many forms of social data are increasingly being georeferenced, distributed, and
archived. This opens up new avenues for the exploration of these data and allows
for new ways of analyzing them, when combined with remotely sensed data.