3. DATA ACQUISITION AND METHODS
The evaluation of the multiseasonal RADARSAT datasets
(May, 15th and October, 23rd 1996 with incidence angles
between 36° - 42° and 30° - 37°, respectively) was performed
according to the following steps: pre-processing of the scenes to
reduce speckle noise, scaling of the data from 16 to 8 bits real
integer and extraction of the backscatter values (linear gamma
values) from previously selected sample areas. The digital
number that is assigned to a specific pixel in the radar image is
related to the return signal (echo) from a corresponding ground
cell at the position (x, y). It is also related to the radar
backscatter cross section (o^) of the cell, which is defined as the
ratio between the incident by the return energy to each cell
(Yanasse, 1991). Another definition related to backscattering
that is often used at radar images to describe the reflectivity of
an area is a “scattering cross section per unit project area" (y^)
which is defined as the radar cross section divided by the area,
normal to the direction of the propagation through which the
surface area is illuminated, that is y° — o^ / cosq.
The TM/Landsat scenes processed by IHS transforms were
resampled and merged to RADARSAT images using a cubic
convolution interpolation, with an error of fitting « 1 pixel) to
identify the different patterns and to locate the sample areas.
The field survey was made concomitant to the RADARSAT
Scene acquisitions (May and October) in order to obtain the
physiognomic-structural ^ characteristics of primary and
secondary forest and also the conditions of the pastures. In those
areas covered by primary and secondary forest (18 samples)
different parameters were estimated: DBH, height, crown cover
percent as well as the botanic species identification. All
individuals with DBH > 10 cm for primary forest and DBH > 3-
5 cm for secondary forest, at samples areas of 10,000 m? and
2,600 m? respectively, were measured. The size of sample plots
was limited by the time available for measurements, however
the sample size was enough to include representative diameter
classes, as well as floristic diversity in primary and secondary
forests. The estimation of biomass values was modeled by DBH
and height into the following allometric equations for primary
forest (Brown et al.,1989):
Y = 0.044 . (DBH? . height)?" ? and
Y =- 2.17 + 1.02 In(DBH) + 0.39 In height
according to Uhl et al.(1988) for the secondary forest. Diagrams
of the dispersion were made to know the relation between
biomass values and seasonal Radarsat backscatter values. In
parallel, structural profiles were made through primary and
secondary successions, in order to analyze and explain the
influence of dendrometric parameters on the behavior of radar
data. A qualitative evaluation of the pasture conditions at both
RADARSAT acquisition dates (begin of the dry and the wet
season) was also made, based on the backscatter changes.
4. RESULTS.
The joint investigations of landscape dynamics in the area under
study with RADARSAT and TM/Landsat scenes, support by
ground control, were the identification of the following
vegeiation cover types: primary forest with and without
bamboo, secondary forest (initial, intermediate and advanced
stage of succession). Pasture, burned and deforested areas were
also identified. Based on those information obtained during
field surveys related to sampling areas (18) of different types of
vegetation cover, it is possible to show the behavior of
RADARSAT backscatter (linear gamma values) for different
seasonal conditions (Figure 2)
Gamma Mean Values
P
P
P
P
PFB1
PFB2
PFB3
InSS1
InSS2
ItSS1
ItSS2
ItSS3
ASSI
ASS2
ASS3
ASS4
ASSS
[—e— May, 15 th —B— October, 23th |
InSS - Initial Secondary Sucession
ItSS = Intermediate Secondary Sucession
ASS = Advanced Secondary Sucession
PF = Primary Forest
PFB = Primary Forest with Bamboo
Figure 2. Diagram of RADARSAT backscatter from primary
and secondary forest samples for different seasons.
Generally speaking, the highest amplitude of backscatter values
(y°) for different seasons, was observed in areas covered with
primary forest with bamboo and secondary forests in growth.
Those areas of primary forest with bamboo (Figure 3) have a
relatively low biomass (76.78 to 113.12 ton/ha), which explains
the high density of this species, where there is a high
competition for the occupation of physical space vis-à-vis
woody species that are commonly found in primary forests. The
height of individuals of the genus Bambusa constitutes the
upper canopy of the forest, presenting a significant spectral-
textural response at satellite data, specially those ones that
backscattering mechanism of the interaction is classified such as
“direct scattering in the crown” (Dobson et al., 1992). The
phases of growth and senescence of these bamboo communities
are determinant for the variations of backscatter found.
Figure 3. Typical features inside a forest with bamboo species
(Guadua weberbaueri Pilger).
528 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
| —&"ELI
Fig
At ]
plot
The
sen:
bior
Camma Mean Valnaa
Neomme Moan WVaalaan
Figu
abo
Succ
