Methods
nts from
ation at
980. A
ehaviour.
Chemical
n, 1985.
RR-Based
Journal of
|.
ent of a
cal Fluid
Physics,
ients with
riment. J.
1993. An
tion of
Between
the GF8
., 31, pp
THE USE OF GAIN CONTROLLED DMSP/OLS DATA TO MAP URBAN/SUBURBAN AND INDUSTRIAL
LAND CONVERSION IN THE UNITED STATES
Imhoff, M. L. - NASA Goddard Space Flight Center, Greenbelt, MD, USA, W. Lawrence and D. Stutzer - Bowie State
University, Bowie, MD, USA, G. Petersen, and E. Nizeyimana - The Pennsylvania State University, University Park, PA,
USA, and C. Elvidge, University of Nevada, Reno, NV, USA
KEY WORDS: DMSP/OLS, Gain Control,
Urban/suburban, Land Conversion, United States
ABSTRACT
Nighttime data sets from the Defense Meteorological
Satellite Program's Operational Linescan System
(DMSP/OLS) have been used to map the location and extent
of human settlements through the detection of "city lights".
Data products from the DMSP/OLS have been used in their
raw form and in the form of composites produced by
compiling multiple data sets:over a period of time. The full
utility of DMSP for urban mapping applications, however,
has been limited by the differences in OLS sensor gain used
between orbital data acquisitions. The unrecorded variation
in gain settings causes differing degrees of image blooming
for a given urban area which cannot be adequately rectified
between data takes. This study employs DMSP/OLS data
collected by the U.S. Air Force and NOAA/NGDC using
gain control to limit sensor saturation over urban areas and
provide light intensity image data of urban, suburban, and
industrial areas, in the United States. The use of gain control
allows for the classification of lit area based on measured
light intensity values and opens the door to a whole new
field of remotely mapping and classifying human settlements
and assessing their impact on the biosphere.
INTRODUCTION
Recently, the phenomenon of urbanization is of growing
interest to the science community. Human population
growth and the trend toward urbanization has pushed urban
land transformation issues into the arena of global climate
change and economic and biological sustainability (Kates et
al., 1990; Ehrlich and Wilson, 1991; Raven, 1991; Daily
and Ehrlich, 1992; Ehrlich and Ehrlich, 1992).
Historically, however, measuring the extent of urbanization
using conventional methods has been problematic even in
the U.S. where modern census procedures are used.
DMSP/OLS Images of Earth's Inhabited Places
Satellite remote sensing can be a useful alternative and
corroborative technology for measuring and monitoring the
location and extent of urbanization. Nighttime images of
Earth's cities acquired from the Defense Meteorological
Satellite Program's Operational Linescan System
(DMSP/OLS) provide a dramatic picture of urbanization
through the detection of city lights. Early work with this
data demonstrated its potential as an urban mapping tool,
(Croft 1978, Welch 1980, and later by Kramer 1994) and as a
means for relating lighted area and human habitation to
energy consumption (Welch and Zupko 1980, and later by
Elvidge et al. 19972). The high contrast between lighted
and unlighted lands, provided by the image data, and the
sensor's moderate spatial resolution (2.7 km) make it an
potent choice for identifying areas where significant human
activity is being carried out.
The National Oceanic and Atmospheric Administration's
National Geoscience Data Center (NOAA/NGDC) has
produced two types of DMSP/OLS data suitable for urban
applications: 1) the "stable lights" or temporal image
composites and 2) gain controlled data designed for urban
imaging. Both types of data are available for the U.S.
(Elvidge 1997b).
The "stable lights" data set is a composite of many images
collected over time. It effectively eliminates ephemeral light
sources such as lightning and fires to provide an image of
consistently lit areas on the Earth's surface. This product
has been shown to be useful for mapping urban areas,
predicting energy use, and for assessing the impact of land
transformation on soil productivity (Imhoff et al. 1997a,
1997b, Elvidge et al. 1997a, Sutton et al. 1997). The stable
lights product is valuable because it eliminates ephemeral
light sources and provides a means of identifying stable
developed areas in the imagery. However, since the gain of
the sensor was not controlled for urban applications, most
urban light emissions saturated the sensor making
relationships between image DN value and radiance level
impossible. Also the saturation effect has a degrading effect
on the 2.7km resolution of DMSP/OLS through blooming.
The blooming created by fluctuating gain settings
(especially at the high end) and the compositing process can
create problems in finer scale applications of the data.
The gain controlled DMSP/OLS data, on the other hand, is
collected using gain settings that are adjusted so that only a
very few large urban areas saturate the sensor while still
being sensitive enough to detect the smaller towns and
inhabited places. The gain controlled data allows for a more
methodical analysis of the data and the ability to translate
the image digital numbers (DN’s) to radiance values in
watts (Watts/cm?/sr/pm). Density slicing the gain controlled
data along the radiance axis has meaning relative to light
emission intensity on the ground and controls blooming.
The gain controlled data used in this analysis were collected
in March, 1996.
METHODS
Thresholding the Gain Controlled Image Data
Prior studies using the stable lights composite product
(data collected from 1994 to 1995) showed that a spatial
integrity determinant could be used to reduce blooming and
create an image product which very accurately mapped urban
area in the U.S. (Imhoff 1997a). Accuracy was assessed by
comparing the lit area defined by the DMSP product to the
area of urban and urbanized land defined by the 1990 U.S.
Census. Results for the “stable lights” product showed an
overall difference between lit area (after the threshold) and
the census estimate of urban lands of only 5%. However, on
a state by state comparison, there were large differences
between the census and the stable lights data depending on
the spatial patterns of development and the above mentioned
limitations of the stable lights product.
For the gain controlled DMSP/OLS data set of the U.S., we
decided to make a similar assessment of the imagery by again
comparing it to the 1990 census data of urban and urbanized
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 435