ind 
cal 
ind 
ind 
a) 
ace 
net 
3. CONSTRUCTION OF DATABASE 
AND PREPROCESSING 
3.1 Construction of database 
3.1.1 Prototype database for hydrological 
monitoring data: Water temperature is the basic 
elements of hydrological data. We obtained information 
on its attributes and sample data were collected by 
questionnaire from many monitoring stations in Japan. 
Also, the sample data were compiled and formatted by 
generation of these attributes from other databases such 
as the public use water quality datafile and Fisheries 
Information Center (Fig. 2). 
Cx TE = x 
PASS 
  
  
  
  
  
  
  
  
  
  
Fig. 2 Distribution of monitoring points of 
water temperature in and around the 
Japan Islands. 
3.1.2 Prototype database for meteorological 
monitoring data in Japan: We obtained 
attributes from about 130 meteorological monitoring 
stations and their sample data using questionnaires. 
The meteorological prototype database was constructed 
in 1994. We were able to exclude noisy data for 
evaluating the circumstances of the monitoring station 
using height and distance from natural and artificial 
features, which affect the data (Fig. 3). 
  
  
  
  
  
  
Fig. 3 Output of retrieval system of the 
meteorological database. 
885 
  
Fig. 4 Output of the results of rectification 
of NOAA imagery for Hokkaido, Japan. 
3.2 Improvement of rectification system for 
NOAA AVHRR imagery. 
À rectification system for NOAA imagery based on 
Sun's algorithm (Sun,1989) was elongated to allow 
efficient data retrieval from rectifiïed NOAA image in 
TIMS (Fig. 4). 
3.3 Cloud screening 
Cloud screening is necessary to reduce the estimation 
error in statistical processing and to obtain the 
amplitudes of both the earth's surface temperature (Ts) 
and net radiation. A cloud screening subsystem was 
developed based on i) a color composite method 
(Utsunomiya ef al, 1990; Bellec and Gleau, 1992), and 
ii) the PCA/ clustering procedure "PCTSMC" 
(Gallaudet and Simpson,1991) using brightness Ts 
derived from NOAA Channels 3, 4 and 5 (Fig. 5). 
These procedures were evaluated using NOAA daytime 
imagery of an area from Hokkaido to  Sanriku 
(northeastern Honshu). The results of both procedures 
were similar, although the latter overeliminated the 
cloud-masked areas (Fig. 6). 
  
Fig. 5 Cloud screening based on the color 
composite method of NOAA imagery 
in northeastern China (after Utsunomiya 
et, al., 1990). 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996