eli
410
minimum air temperatures observed for urban and rural
locations. The satellite-derived vegetation index data of a
monthly composite were lincarly related to the difference in
observed urban and rural minimum temperatures, explaining
37% of its variation (versus 15% using radiant surface
temperature). Thus, during the growing season and in
well-vegetated regions, a vegetation index is a useful
parameter for assessing the UHI effect.
In addition to land cover parameters derived from the
AVHRR, Gallo er al. (1995) found that “city lights” data
(Elvidge er al., 1997) derived from the 0.6-0.9 um band
of the Operational Linescan System (OLS) of the U.S. Air
Force Defense Meteorological Satellite Program (DMSP)
appeared useful in distinguishing between urban and rural
locations. The “city lights” data set offers the prospect of
periodic and objective assessments of urban expansion at
a continental-scale.
The general objective of the paper is to provide an
overview of the application of data and products available
from the AVHRR, Landsat MSS and DMSP-OLS sensors
to studies of urban climate. A more specific objective is
to demonstrate how the data from these sensors can be
used to assess whether a given meteorological station can
be characterized as "urban" or "rural." An objective
method for determining whether‘à station is urban or rural
would bc beneficial for the assessment of global climate
change, as the influence of urban stations could be
extracted from future analyses of temperature trends.
DATA AND METHODOLOGY
Data Analysis
The Dallas-Fort Worth (DFW), Texas, USA, metropolitan
arca was selected for this analysis. The population
associated with this metropolitan region has increased
from 2.3 million in 1970 to an estimated 4.4 million in
1995 (U.S. Bureau of the Census, 1997) . AVHRR and
DMSP-OLS data were sampled for 7 climatological
stations (Table 1) included in the DFW region. These
stations are part of a larger network of stations throughout
the USA being utilized to examine the heat-island effect
(Gallo et al., 19933). The reported AVHRR and DMSP-
OLS values are the mean values for 3X3 km? “windows”
centered on the location of the climatological stations,
unless otherwise noted.
AVHRR-Derived Vegetation Index
Normalized Difference Vegetation Index (NDVI) values,
computed (equation 1) from the visible (C1; 0.58- 0.68
pm) and near-infrared (C2; 0.72 - 1.1 um) data acquired
by the AVHRR as
NDVi = (C2-C1)/ (C2 + C1), [1]
were obtained from two products for use in this analysis.
The NDVI, like most other vegetation indices, exploits the
difference in the reflectance of vegetation in the near-
infrared wavelengths compared to visible wavelengths.
Multiple scenes for a given time period (typically 10-day,
14-day or monthly) are composited by retaining, on a
pixel-by-pixel basis, the maximum observed value of
NDVI during the interval. Composited NDVI images
minimize the probability that pixels with cloud or haze
contamination (which result in lower NDVI values) will
remain in the image.
Fourteen-day composited NDVI and individual channel
AVHRR data are readily available for the Conterminous
USA at the 1-km resolution in a Lambert Azimuthal
Equal Area (LAEA) projection. NDVI values were
extracted from a CD-ROM product (USGS, 1991) for two
14-day composite intervals of 1991. The data of the two
14-day composites were then further composited over a
28-day interval (5 July through 1 August 1991).
AVHRR-Derived Radiant Surface Temperature
Radiant surface temperature (Tsfc) can be approximated
from calibrated thermal IR data of channels 4 (T4; 10.3 -
11.3 um) and 5 (T5; 11.5 - 12.5 um) of the AVHRR
sensor for a surface emissivity of 1.0 as suggested by
Price (1990) as
Tsfc = T4 + 3.3(T4-T5) . (2)
The T4 and TS data associated with the NDVI values
composited over the 5 July through August 1991 interval
were used to compute Tsfc. The Tsfc of this interval was
selected for analysis to permit direct examination of the
relationship between Tsfc and NDVI at the climate
stations.
DMSP-OLS-Derived “City Lights" Data
The DMSP-OLS data have historically been used to
monitor the global distribution of clouds and cloud top
temperatures. However, the data acquired in the
visible/near-infrared band (0.6-0.9 um) at night under
cloud-free conditions can be used to identify light emitted
at the surface of the earth. Gallo e: al. (1995)
demonstrated that this data might be useful for
identification of urban and rural locations as it primarily
identifies the light associated with urban locales (e.g.,
“city lights”). However, caution must be used when these
data are used for identification of urban areas as, on any
single night, the burning of forests (e.g., burning biomass)
and fossil fuels could be misinterpreted to represent the
location of an urban areas.
Two DMSP-OLS data sets were used in this analysis. The
first DMSP-OLS data product was based on a composite
of 231 orbits that were used to identify stable (e.g., urban
areas) from non-stable (e.g., burning biomass) light
sources. The orbital data were mapped to a |-km
projection (Goodes Homolosine) to match the globally
available 1-km AVHRR data. The DMSP-OLS data were
provided by the National Geophysical Data Center and
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998