increased
ity of two
reservoir,
he type of
(
x
>
und in the
Pistia sp.
(floating
us sp. Will
ine roots,
to 15 cm
the most
ith stands
of 1 to 2
Since the
> areas arc
the water
^ b) CAN ue -
ure 2. a) Scirpus sp.; b) Eichhornia sp..
3. METHODOLOGY
3.1 Data Acquisition
From April 14 to 16, 1992, the Brazilian National Institute for
Space Research (INPE), in co-operation with the European
Space Agency (ESA), the Canadian Centre for Remote Sensing
(CCRS) and Centrais Elétricas do Norte (ELETRONORTE)
carried out an experiment in the Tucurui reservoir, to assess the
feasibility of using SAR data to map aquatic vegetation. The
campaign included acquisition of multiviewing and
multipolarization data, in the Nadir mode, taken by the Convair
580 SAR C-band System, from CCRS, over the inlets named
Pucuruí and Repartimento in the Tucurui reservoir. The swath
width of the Nadir mode is 20 km and the incidence angles are
between 20°-74°. Details of the campaign are found in
Wooding and Zmuda (1993) and Novo et al. (1995).
The characteristics of SAR data used in this investigation are
presented in Table 1. According to Hawkins and Teany (1993),
the data acquired during the SAREX 92 mission were submitted
to relative calibration. The calibration constants were
empirically determined using corner reflectors deployed near
Ottawa, Canada. These data were collected just before and soon
after the mission, in order to assess the stability of the system.
ERS-1 calibration data were then used to adjust the SAREX
data.
Frequency 5.3 GHz
Wavelength 5.66 cm
Polarization HH, VV, HV, VH
Mode Nadir
Incidence angle of test area 38° - 50°
Pixel size 4m x 4.31m
Resolution 6m x 6m
Number of looks 7
Table 1. SAR characteristics.
The whole calibration equation suggested by Hawkins and
Teany (1993). The INPE processed images have as output the
amplitude values (A), given by:
529
Fcal( j)- Fcal
A - JDN? (ij) - DN,?(j) id 2) (1)
where:
- DN(ij) is the pixel value at row i and column 7;
- DN,(j) is the noise value at column j;
- Fcal(j) is the calibration factor at column j.
- Fcal is the mean value of the calibration factor.
The equation used to convert the digital numbers (DN) into
backscatter values (6°45) was:
6° = 10 log,, A> + Fcal Q)
Aerial photographs at the scale of 1:10 000 were acquired
concurrently to the SAREX 92 mission and used as ground
truth. A digital mosaic of 12 aerial photographs was registered
to the SAR image. The mosaic was also used to set the limits of
selected samples from different species. The number of samples
(n) from each class was proportional to their occurrence in the
reservoir. The classes studied are: Open water, Dead tree,
Forest, Eichhornia sp., and Scirpus sp.. For every sample the
mean and the standard deviation of the digital number were
computed in the four polarizations. Each mean digital number
was converted into 6° through equation (2). Values of Fcal
for each polarization are presented in Table 2.
HH - -48.660118
VV = -48.40868
HV = -45.51140
VH = -45.634056
Table 2. F cal values.
The coefficient of variation was calculated for all classes using
the digital numbers.
3.2 Correction Factor
The data presented cross-polarization dB values not compatible
with that reported in the literature and neither consistent with
radar theory in C band. The cross-polarization backscattering
dB values were higher than the like-polarization. Hence, a
correction factor (k) was computed. It is formulated as follows:
PA FH 159 13 106/65 26-2755 giqp (3)
where,
N
ZH; + VV;)
À vH-vv = i5 = 1.3 dB (4)
N
were Avy VA is the mean variation in all classes
distinguished from SAREX data. This mean variation
represents how large are HV data compared to the VV data.
S- VV-HV = 66dB (5)
where VV and HV are the mean values, in C band, for the
following classes: shrubs, short vegetation and grasses, at an
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996