Novo, Evlyn 
  
The main disadvantage of the WFI/CBERS camera as compared to other remote sensing systems currently in use for 
water quality assessment is the reduced number of wavelengths (2 bands) and the spatial resolution (258 m by 258 m at 
nadir). The reduced number of wavebands theoretically is not a major problem if the interest is to map the three major 
water types found in the Amazon region. The literature shows that those water types differ significantly in the total 
suspended solids. White water have extremely high total suspended solids whereas black water have extremely low and 
clear water have intermediate values (Junk and Furch, 1985; Richey et al, 1986). Several authors reported high 
correlation between red and near-infrared reflectance and total suspended solids at concentrations larger than 10 mg/l 
(Lathrop and Lillesand, 1989, Novo et al., 1991). 
The spatial resolution, therefore, seems to be the most restrictive feature of the WFI/CBERS camera and must be 
investigated before WFI data can be recognized as a new tool for monitoring water types in the Amazon basin 
floodplain. 
  
  
  
  
  
  
  
  
  
Center Wavelength 654 nm 834 nm 
Bandwidth at 0.5 peak transmittance 62 nm 126 nm 
Band width 630 to.690 nm 770 to 890 nm 
Swath coverage 890 km 890 km 
Radiometric Range Min — 0.01 W/m/ .sr Min = 0.02 W/m” sr 
Max=9.5 W/m’ ‚sr Max = 15.1 W/m°.sr 
Radiometric Levels 256 levels 256 levels 
Spatial Resolution 258 m x 258 m 258 m x 258 m 
Frequency of acquisition 5 days 5 days 
  
  
  
  
Table 1 — WFI/CBERS camera features (Source: http://www.inpe.br/programas/cbers/portugues/index.html) 
The present study was proposed to assess the potential of the WFI camera to gather information about the spatial 
distribution of water types in the Amazon River Floodplain. In this study, 29 TM-Landsat images assembled into a 
digital mosaic (Shimabukuro et al., 1998) were used to simulate the spatial and spectral features provided by the WFI 
camera. 
METHODS 
A full description of the methodology used to build the mosaic is found in Shimabukuro et al. (1998). The first step in 
this simulation was to resample band 3 (red) and 4 (near infrared) of the TM-Landsat digital mosaic to 100 m by 100 m 
and to 258 m to 258 m spatial resolution mosaics. The resampling was performed using a nearest neighbor algorithm to 
reduce radiometric degradation of the original 30 m by 30 m resolution TM data (Richards, 1995). The rational for 
producing a 100 m by 100 m mosaic is the possibility of resampling WFI images to that resolution using data fusion 
procedures and image restoration techniques (Richards, 1995). 
In spite of the radiometric rectification developed by Shimabukuro et al. (1998), several mosaic scenes of band TM 3 
were heavily affected by cloud cover. To overcome this problem, band TM 4 was used to generate a water body mask. 
This mask was then used to exclude the land (mostly covered by clouds and cloud shadows in band 3). The masked 
bands (TM 3 and TM 4) were then added and linearly scaled to 256 digital counts to produce a single band sensitive to 
changes in particle concentration (Novo et al., 1991). At both resolutions, the minimum and maximum number of 
classes was set as 5 and 10, respectively. The resulting spectral classes were examined and combined according to the 
range of digital levels of known water body types such as Rio Negro (Black Water Type), Tapajós (Clear Water Type), 
and Solimóes (White Water Type). 
The proportion of water types was determined at selected Amazon River reaches. Changes in the proportion of the 
various water types were then graphically compared to examine the effect of spatial resolution on the water type 
mapping. 
RESULTS 
Figure 1 represents the color composite of simulated WFI red and near infrared bands as follows: Red Band (Red and 
Blue), Infrared Band (Green). In spite of the coarse resolution, the simulated WFI composite allows the identification of 
two major water types in this reach. White Water, seen as magenta, is responsible for the strong reflected radiance in 
  
1028 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 
  
  
Jg 
(