International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
analysing multitemporal SAR images. The applicability and 
efficiency for updating of various spatial data was evaluated. 
2. MATERIALS AND METHOD 
2.1 Study Site and Image Data 
JERS-1 SAR images were used in this study. The JERS-I 
satellite was operational between 1992 and 1998 and it had a 
repeat cycle of 44days. JERS-1 has already finished its mission, 
and the launch of a successor satellite (ALOS) is planned in 
2004. This study is setting its sights on the practical use of 
ALOS/PALSAR data, and the method will be applied to 
ALOS/PALSAR after development based on JERS-1/SAR. The 
JERS-1/SAR sensor is quite suitable for the purpose of this 
study, because of the sensor's stable periodicity of data 
acquisition. The periodicity of data from these sensors is about 
1.5 months, which is satisfactory when compared to the 
frequency of updating spatial data (1-5 years) in general. 
Kanagawa prefecture, Japan, was selected for the study site. 
The reason for this was that Kanagawa prefecture has 
appropriate characteristics as follows: a) There are various 
geographical characteristics (e.g. urban areas, agricultural areas, 
mountainous areas and water areas), and b) The area is located 
on the urban fringe of Tokyo. Figure 1 shows a comparison of 
data acquisition ratio between JERS-1/SAR and JERS-1/VNIR. 
It is clear that optical sensor image has a limitation for 
periodical monitoring. 
2.2 Method 
There are two types of features on the land surface. One is a 
changing features and the other is an unchanging features. The 
changing features can be further divided into stationary changes 
and nonstationary changes. The change in the agricultural area 
is an example for stationary change. The agricultural area 
changes its land surface in a certain cycle according to the stage 
of farming, such as seeding, growing, and harvesting. These 
changes are stationary as the land use does not change, and they 
should not be a trigger of updating for land use data. On the 
other hand, nonstationary change occurs randomly and 
suddenly most of the time. Nonstationary change can be 
explained by taking urban development and natural disasters as 
an example. Most of nonstationary changes require data 
updating, while stationary changes often do not. The rule for 
updating spatial data depends on the characteristic of each 
thematic data. The rule is decided mainly by the stationarity of 
the land surface change. Therefore, these two types of changes 
have to be considered as completely different changes. The 
method focused on the role of a support tool for making 
decisions as to whether or not the spatial data should be updated. 
Generally, changes of the feature are extracted by subtraction of 
two data sets having different time stamps. However, these 
methods based on pairs of data cannot tell whether the changes 
are stationary or nonstationary. Stationarity of change can be 
observed correctly only when data is archived in for a long 
enough time, periodically, and frequently enough. This 
  
  
30% 
20% 
area without cloud-cover 
  
10% 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
92 1993 
1994 1995 1996 1997 1998 
  
0% 
10% 
20% 
30% 
40% 
60% 
  
70% 
80% 
JERS-1/SAR 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Figure 1. Comparison of data acquisition ratio between JERS-1/VNIR and JERS-1/SAR 
Inter) 
  
requi 
not r« 
sense 
most 
The 1 
of ^ 
and s 
by cc 
1). TI 
In th 
series 
and : 
disad 
make 
reduc 
resoli 
to mi 
maxi 
PCA 
imag: 
multi: 
Coltu 
of the 
of fe: 
on th: 
chang 
fluctu 
meth« 
seasoi 
patter 
inside 
using 
the N 
methc 
as a te