e orbital
or 10/96.
bit of the
data was
ges.
t “city
ASP/OLS
(d it with
n errors,
er bodies
rban land
4] and
TM data
OS Data
quired to
that were -
189/039,
M scenes
WRS II].
ication of
type land
gital Soils
et for this
ion, is of
database
JNESCO,
CA USA]
he image
Ground
egistering
orbit data
UNESCO
ta as they
n the data
| common
Landsat
area was
X 120 km
N31? 10'
N29» 38"
and most
nation of
ty lights"
ducts -
a manual
tal values
ble. This
it desert,
DMSP/OLS Single Orbits
10/77 06/84 07/86 02/87 01/88 02/90 10/96
Date of Acquisition
Figure 1. Historical trend in urban land cover, Cairo/Nile
delta region. 19777-1996 using DMSP/OLS ‘paper’ data and
single orbit products.
ephemeral light and low-level illumination from each of the
images. The process was keyed on the removal of ‘noise’
and the retention of spots of know urban population centers.
Thresholding was stopped at an upper limit when the
population centers disappeared or were reduced in size. Total
area was calculated on a summed pixels within class basis
and compared year-to-year.
DMSP/OLS single swath products - A manual
thresholding process was used for this digital product that
followed the protocol developed for the scanned non-digital
products. Total area was calculated based on pixel size and
sum.
2.32 Landsat Images: The four 1987 Landsat Thematic
Mapper and four 1976 Multispectral Scanner scenes were
each mosaiced in the registration process. An unsupervised
classification routine was used to create clusters of pixels
with similar spectral characteristics [ENVI, Isoclus, PCI,
Canada]. The clusters were identified manually using the TM
high resolution images in natural and false-color near IR
representations, and other collateral data as available.
Clusters were combined into six cover types: urban, water,
vegetation, bare soils, desert, and other [<2%]. This study
area was approximately 14,400 km? [see 2.3 above].
2.33 FAO Soils with *City Lights" Urban Land
Cover: A spatial intersection of the FAO Soils classes was
made with the "city lights" urban land cover data set. We
used a popular geographic information system [GIS,
ArcInfo, ESRI, Redlands, CA USA] for this operation.
Percentage of each of eight soil types plus water under “city
lights” or urban land use were calculated for the entire
country of Egypt [see 2.3 above].
3. RESULTS AND DISCUSSION
3.1 Urban Land Cover Change
3.11 DMSP/OLS-based Urban Land Cover
Change - Both ‘historical’ [1977-1990] non-digital
sources of nighttime images and a single orbit digital
DMSP/OLS image [1996] were used in this analysis to see if
historical trends in urban land cover area could be determined
| igure 2. Historical ‘paper’ product from DMSP/OLS. Note
high gain allows visualization of not only lights from urba
areas, but also clouds and moonlit desert landscape.
for the Cairo/Nile delta area. After the historical data sets
were digitized and ‘thresholded’ to remove ‘noise’ they were
compared year-by-year including the 1996 digital data set.
The results of this comparison [Figure 1] show that it is
extremely difficult to use either the early non-digital
products or the current digital DMSP/OLS products in a
fruitful quantitative analysis. Most of the difficulty lies in
- selecting appropriate digital values that can differentiate
between actual urban land cover classes and other
inappropriate classes. Some of this confusion is due to the
wide range of land use types that can fall under the category
of urban land cover, but much is due to the unknown gain
settings of the instrument which is varied by the DoD to
optimize cloud detection through the monthly phases of the
Moon, not to optimize delineation of urban land cover
types. Using unaltered data, the percentage of urban land
cover varied from 22.8 to 64% of the 14,400 km?2 study area.
After applying a threshold filter to the data, the range was
7.4 to 42.4% of the area under urban land cover. There is no
consistent trend in an area known for its rapid urbanization
and population growth. Neither the ‘historical’ nor the
single orbit digital DMSP/OLS data sets are very useful for
areal studies without further processing that was beyond the
scope of this project.
Some of the problems in interpretation of this data set can
be seen in Figure 2 which shows one of the ‘historical’ data
sets. With the gain in use at the time of acquisition,
extraordinary detail can be seen in land cover detail and
moonlit clouds well beyond just the extent of lighted areas
of human occupation. A more typical image is that shown in
Figure 3, the 1996 digital product of the fraction of a single
DMSP/OLS orbit. The thresholding process is represented
in Figure 4 showing a subset of the digital 1996 image of the
Cairo and Nile delta region in ‘raw’ form, with the ‘best’
threshold, and with only the highest level of illumination or
pixels with an exoatmospheric radiance that saturated the
Sensor.
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 445
