Noernberg, Mauricio Almeida 
  
The data used in this study were acquired during the SAREX 92 mission, which provided an 
unprecedented opportunity to study the behavior of tropical targets using airborne multipolarimetric C band 
SAR data. This band is also available in orbital SAR systems in operation since the 90’s, such as the ERS-1 and 
2, and RADARSAT. The overflights were carried out on April, 14, 1992, comprising five swaths with the C/X 
SAR/CCRS. At the same time, the Bandeirante aircraft from the National Institute for Space Research covered a 
swath over the Pucurui inlet, generating color aerial photographs at the scale of 1:10,000. Simultaneously, a 
team of researchers carried out field work at the Tucurui reservoir, aiming at locating and identifying dominant 
aquatic plants stands. The SAR images were subjected to the correction procedures suggested by CCRS (Canada 
Centre of Remote Sensing) (Hawkins and Teany, 1993), remaining as range images. For more details about the 
SAREX 92 mission in the Tucurui reservoir see Novo et al. (1995). Table 1 presents some features of the SAR 
images acquired during the SAREX 92 mission. 
  
  
  
  
  
  
  
frequency 5.3 GHz 
wavelength 5,66 cm 
polarization HH, VV, HV, VH 
mode Nadir 
average incidence angle 38° to 50° 
pixel size 4m x 4,31m 
resolution 6m x 6m 
looks 7 
  
  
Table 1. Characteristics of the SAR data. 
Aerial photographs at the scale 1:10,000 were used as ground truth, due to the inaccessibility of the 
study area. First, the aerial photographs were converted into digital images, originating a mosaic that was 
registered with the SAR images. The interpolation algorithm used was the nearest neighbor. The SAR images 
were not registered to the topographic sheets to preserve their radiometry. Once the images were registered, a set 
of samples was selected using the aerial photographs as reference. The number of samples from each class was 
based on their incidence in the reservoir. For more details on sampling selection see Noernberg (1996). Table 2 
introduces the sampled classes, the corresponding number of samples, and the number of sampled pixels from 
each class. 
  
  
  
  
  
  
Ground Class Sample Size Number of Pixels 
Scirpus 39 155.176 
Eichhornia 8 14.973 
Typha 8 3.496 
Salvinia 19 26.607 
Pistia S 4.078 
Heterogeneous 9 69.804 
  
  
  
Table 2. Number of samples and total number of pixels for each of the sampled classes. 
The total number of pixels was used to test the goodness of fit of the data to different statistical 
distributions. Two general types of distributions were tested: (a) those originated from the multiplicative model 
(Square Root of Gamma and Amplitude G0) (Yanasse et al., 1994; Vieira, 1996); and (b) those not originated 
from the multiplicative model (Normal, Log-Normal, Weibull) (Yanasse ety al. 1994). The test used was the chi- 
square goodness of fit (Sokal and Rohlf, 1969), which is based on the value of the x. statistics defined as the 
sum of the quotients deviations (observed values minus expected values under an assumed distribution) squared 
over expected frequencies. Under certain conditions, the x statistics follows a x distribution. The number of 
degrees of freedom depends on the hypothetical distribution. The software used to test the goodness of fit to the 
various distributions was developed at INPE (Vieira, 1996). 
The software provides the p-value of goodness of fit tests, which indicates the probability of obtaining a 
value of Y as high as (or greater than) that observed, assumed a given distribution. If the defined level of 
significance is &, then any p-value smaller than leads to the rejection of the null hypothesis. The statistical 
  
1012 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.