XXXIX-B8, 2012
el for prediction of
a Tropica, 96 (2-3),
Li, B., 2010. Linear
lata for monitoring
inds. International
2-4333.
, Koch, M. and
and temporal
literranean wetland
ournal of Remote
2. Spectral mixture
g Landsat TM and
f Remote Sensing,
e remote sensing of
nt, 10 (5), pp. 381-
tin, S., Scheer, G.,
1 the Santa Monica
al mixture models.
67-279.
P.E., 2007. VIPER
ty of California at
ccessed: March 8,
J.B., 1993. Green
d soils in AVIRIS
>), pp. 255-269.
| of gross primary
ding with MODIS
^ Observation and
J.M., and Novo,
vegetation: theory
> and Assessment,
.G., 1982. On the
er data. Canadian
d Hilker, T., 2011.
ge adoption using
y. Remote Sensing
X.H., Jiang, Q.W.,
The public health
China — then and
ek, J.G., 2010. An
eflectance fusion
emote Sensing of
AUTOMATED CONSTRUCTION OF COVERAGE CATALOGUES OF ASTER
SATELLITE IMAGE FOR URBAN AREAS OF THE WORLD
Hiroyuki Miyazaki “*, Koki Iwao °, Ryosuke Shibasaki *
* Center for Spatial Information Science, The University of Tokyo, Japan - (heromiya, shiba)@csis.u-tokyo.ac.jp
^ National Institute of Advanced Industrial Science and Technology, Japan — iwao.koki@aist.go.jp
Commission VIII, WG VIII/8
KEY WORDS: coverage catalogue, urban area, gazetteer, ASTER, metadata database
ABSTRACT:
We developed an algorithm to determine a combination of satellite images according to observation extent and image quality. The
algorithm was for testing necessity for completing coverage of the search extent. The tests excluded unnecessary images with low
quality and preserve necessary images with good quality. The search conditions of the satellite images could be extended, indicating
the catalogue could be constructed with specified periods required for time series analysis. We applied the method to a database of
metadata of ASTER satellite images archived in GEO Grid of National Institute of Advanced Industrial Science and Technology
(AIST), Japan. As indexes of populated places with geographical coordinates, we used a database of 3372 populated place of more
than 0.1 million populations retrieved from GRUMP Settlement Points, a global gazetteer of cities, which has geographical names of
populated places associated with geographical coordinates and population data. From the coordinates of populated places, 3372
extents were generated with radiuses of 30 km, a half of swath of ASTER satellite images. By merging extents overlapping each
other, they were assembled into 2214 extents. As a result, we acquired combinations of good quality for 1244 extents, those of low
quality for 96 extents, incomplete combinations for 611 extents. Further improvements would be expected by introducing pixel-
based cloud assessment and pixel value correction over seasonal variations.
1. INTRODUCTION
Urbanization has been a main concern for regional and global
environmental change (Foley et al., 2005) and socio-economic
problems (Angel, Sheppard, and Civco, 2005). Various kinds of
studies have used satellite-derived global urban area maps to
evaluate critical aspects of urbanization for global
environmental change, such as size, scale and form of cities and
conversion of land cover. The studies using global urban area
map had provided valuable information of urbanization
especially for less documented regions. As the studies on
urbanization progressed, however, 1-km spatial resolution of
global urban area map have gotten obsolete for measuring
spatial structure of urban area in fine scale (Angel, Sheppard,
and Civco, 2005) and for modelling land use conversion with
socio-economic variables (Nelson and Robertson, 2007).
Developing high-resolution urban area map would be key issue
for promoting new insights on urban dynamics. Several studies
had developed urban area map only for their cases using high-
resolution satellite images (e.g. Landsat, Terra/ASTER,
IKONOS and Quickbird) and shown valuable outcome;
however mapping urban area using high-resolution satellite
images involves considerable time and labour cost. It prevents
comprehensive and comparative studies on urban dynamics on
world’s cities.
Extents of urban area are often broader than coverage of a
satellite images. Therefore, you are required to collect satellite
images for a city when you conduct satellite-based analysis of
urban area of the city. We regard ready-to-use collection of
satellite images for every city will promote comprehensive
studies of urban area using satellite images. In this paper, we
present a method for constructing catalogues of satellite images
of cities of the world. We also present a result of the method
applied to metadata database of ASTER satellite images for
3372 cities of the world.
2. METHODOLOGY
The procedure for selecting satellite images for a city from
millions of scenes of ASTER/VNIR is as following: defining
inclusion extent, within which we constructed a mosaic of
satellite images for a target crowd of cities (COC); sending a
spatial query to identify scenes covering the inclusion extent;
and assigning the orders to reduce cloud contamination. Here,
we describe each step.
2.1 Defining urban extent of cities
For the urban area mapping, we had to define the spatial extent
to be mapped. Fortunately, many existing maps of broad scale
have already spatially indexed the cities of the world with
geographical coordinates. Among them, we employed GRUMP
Settlement Points (GSP; http://sedac.ciesin.columbia.edu/gpw/),
a global gazetteer of populated places, as a primary index of the
cities. The first reason of using it is that it had an almost
complete coverage for the cities of more than 1000 population.
Second, it had been manually associated to geographical
coordinates of the cities. This direct human input is
indispensable for accurate association of place names with
geographic data because insufficient information from the
source prevents automatic matching (Doerr and Papagelis,
2007). Third, GSP includes the estimated population of each
city, which can be used to order priority of the urban area
mapping.
