stanbul 2004 
THE DEVELOPMENT OF THE METHOD FOR UPDATING LAND SURFACE DATA 
BY USING MULTI-TEMPORALLY ARCHIVED SATELLITE IMAGES 
Y. Usuda *, N. Watanabe, H. Fukui? 
* Graduate School of Media and Governance, Keio Univ., 5322 Endo, Fujisawa, Kanagawa, Japan 
- (usuyu, hfukui)@sfc.keio.ac.jp 
b wie . . . . — . 
College of Humanities, Chubu Univ., Kasugai, Aichi, Japan- nov@isc.chubu.ac.jp 
ICWG II IV 
KEY WORDS: Multitemporal, Satellite, SAR, Change Detection, Updating, Land Cover, Land Use, Vegetation 
ABSTRACT: 
This study focuses on the problem of "time" affecting spatial data. The most substantial temporal problem for comprehensive use of 
spatial data is “time inconsistency". Reducing *time inconsistency" within individual data, among multiple data, and between data 
and the real world can be achieved by frequent updating. The utilization of remote sensing is an effective method for updating 
various land surface data (e.g. land use, vegetation, soil and geology). The periodicity of usable data acquisition is quite stable for 
Synthetic Aperture Radar (SAR) data, because of its weather independence. The aim of this study was to develop a method for 
frequent updating of spatial data by taking advantage of the stable periodicity of multitemporal SAR images. Firstly, periodical and 
multitemporal SAR images were integrated by time series analysis. Following this process, temporal changes were restructured as a 
change process model and Speckle noise was reduced. The change process model described the trend of land surface changes. 
Secondly, newly acquired SAR image was assimilated into the database to strengthen the stability of the change process model. At 
this time, changes in land surface data were detected by comparing the change in the newly acquired image with the change process 
model. The effectiveness of this method was evaluated through comparison with actual spatial data. It is hoped that change detection 
in near real-time will be achieved using these procedures. 
1. INTRODUCTION can directly affect the results that are calculated from these data. 
It can be quite dangerous to use spatial data without being 
Improvements in Geographic Information System (GIS) aware of such problems. The problem may be worse nowadays, 
technology have expanded the potential of this powerful tool when considering the improvement of GIS: rapid spread of 
for solving complicated issues such as environmental problems, application fields, trends of data integration, and its impact on 
city planning, and disaster management. Issues are understood societies. It would be meaningful to consider the effect of “time 
through the analysis of spatial data in GIS. These spatial data inconsistency” within spatial data as an important problem, and 
are an abstraction of features existing in the real world. In this to evaluate the effect on spatial analysis. The following two 
sense, the accuracy of spatial data may directly affect the approaches can be taken to alleviate this problem. The first 
reliability of the analysis and understanding of the issues. approach is to decrease "time inconsistency" within spatial data, 
Conventionally, existing spatial data, especially land surface and the second approach is to take this inconsistency into 
data (e.g. land use, vegetation, soil and geology) have the account in using the spatial data. It would be ideal to merge the 
following problems. above two approaches into one process. “Time inconsistency” 
within spatial data would be decreased by the former approach, 
(i) Long-term survey may introduce problems of "time and the remaining inconsistency would be extracted and its 
inconsistency" into individual spatial data sets. Features effect on the result quantified by the latter approach. This paper 
within these data cannot be assumed to exist simultaneously. mainly focuses on the former approach, which is implemented 
(ii) Different types of spatial data each have specific dates of as a post-processing step to decrease the “time inconsistency” 
acquisition (time stamps), periodicity, and frequency of of spatial data. 
updating according to the way they are measured. There 
Will be “time inconsistency” among multiple spatial data. The utilization of satellite remote sensing can be considered as 
This suggests that some errors are inevitable in the results a most suitable tool for frequent spatial data updating. Earth 
a from integrated analysis using these data. observation satellites have advantages in periodicity and 
(ii) When the data updating is time consuming (from continuity of spatial data acquisition when compared to other 
measurement to distribution), it may cause ^time ^ monitoring techniques. Optical sensors acquire rich information 
inconsistency" between the spatial data and the real world. in various spectral bands and provide clear images. However, 
In this case, current issues or problems are not able to be optical sensors have a limitation in periodicity for useable data 
considered by suitable data. acquisition. In contrast, Synthetic Aperture Radar (SAR) 
sensors are almost independent of weather conditions, and data 
Most conventional studies on data integration using GIS have acquisition is performed in quite a stable manner. This study 
not referred to the effect caused by “time inconsistency” in developed a method to: 1) detect changes in features (land 
Spatial data. The problem is only dealt with by using data with surface objects), and 2) update land surface data frequently by 
the closest time stamp. “Time inconsistency” within spatial data 
CA 
O3 
~J