roughness. With the end of the reproduction stage, several
bamboo species are characterized by a massive mortality of
population (Janzen, 1976), as observed in the dominant PEI
species, and the results are extensive clearings with a large
volume of dry biomass. Being so, the senescence period can be
clearly identified by very clear tones due to the dominance of
dry materials.
With the mass mortality of bamboo populations, many of these
areas within the PEI are colonized again by other bamboo
species, restarting the dominance cycle. In the area under study,
two bamboo species predominate: Guadua tagoara (Nees)
Kunth (popularly known as taquaruqu) and Chusquea oxylepis
(Hack.) Ekman (popularly known as criciuma). Both species
have associated occurrences at several points within PEI. G.
tagoara presents characteristically developed trunks which use
thorns to climb arboreal individuals and to establish itself at the
forest canopy, while C. oxylepis, with a similar structure as a
climbing plant, forms a closed cover over the canopy. In spite
of the distinct strategies for its dominance, both species
originate a similar structural pattern, characterized by a
discontinuous canopy, low density of arboreal individuals of
medium and large size. Additionally, the superposition of the
reproduction cycles of these species masks its individual
characteristics in the image. In May 2006, when the QuickBird
image was taken, the populations of taquarugu from the area
under study were at the end of the reproduction cycle, while
this time was the height of the reproduction period for criciuma.
The structural similarity of vegetation together with the
simultaneous flourishing of these species, makes it difficult to
establish distinct patterns for them on a QuickBird scene.
The map of land cover generated from the visual interpretation
of the QuickBird image (Kappa = 0.85) allows the evaluation of
class extension, which forms a mosaic composed by vegetation
in different succession stages, and many of them have a
dominant occurrence of bamboos at different life phases. In the
Amazon, each internally synchronized bamboo population,
occupies extensive areas between 10 2 to 10 4 km 2 , which can be
detected with TM-Landsat and MODIS (Nelson et al, 2006)
images during the massive mortality of these sections. At PEI
however sections of bamboo dominance occupy areas between
0.15 to 0.30 km 2 , which cannot be identified with medium
resolution sensors (Araujo et al, 2005), independently of its
phenologic stage.
From the land cover map it is also possible to evaluate the
extension of the bamboo occupation at Intervales. In the region
analyzed, despite the class forest still covers 36% of the total
area, those classes with bamboo dominance come up to 21%.
The class with spaced bamboo sections, due to the border
contact with bamboo dominance, besides the extremely
aggressive habits of these species, has a high chance to be
progressively converted to the class secondary succession with
bamboos. With this inclusion, the percentage of areas with
bamboo is also 36%. A similar inference can be made for the
class advanced secondary succession covering 21% of the area
under consideration.
Despite the relative restricted number of bands at the QuickBird
image, when compared with other products of remote sensing,
the analysis of the spectral responses from different targets
evidences the discrimination between classes of interest, as
observed at Figure 5. Generally the classes present a typical
spectral behavior of vegetation, with low values at bands 1, 2
and 3, referring to the visible section of the electromagnetic
spectrum and high values at band 4, corresponding to the near-
infrared section, which is more adequate for the discrimination
of these forest targets. Similarly to what was observed in other
scales of work (Nelson et al, 2006; Mendoza et ai. 2004), the
layer formed by bamboo leaves in the forest canopy reflects
more intensively, especially when considering the near-infrared
band. We emphasize again the difficulty to discriminate the
class with bamboo dominance in growing stage, which has a
very similar spectral response as the class advanced secondary
succession, despite the floristic and structural differences.
The adequate mapping of this secondary succession with
bamboo dominance in vegetative stage and the understanding of
the phenologic aspects of the dominant bamboo species, is an
important aspect for studies of forest dynamics. These factors
allow the planning to open new clearings, which is of
fundamental importance to understand the landscape of this
region. In this case, the use of a temporal series is fundamental
for such studies, due to a very specific dynamic of this forest
formation.
Figure 6a illustrates a section of bamboo at reproductive stage,
mapped with aerial photography and the corresponding
QuickBird image. The aerial photography was taken in 2000,
probably at the growing stage. Figure 6b shows part of the
phenologic cycle from the bamboo at an advanced
decomposition stage in 2006.
(a)
(b)
Figure 6. Examples of the phenologic phases of bamboo at
aerial photography from 2000 (1) and QuickBird image from
2006 (2): [VB] bamboo during vegetative stage, [RB] bamboo
during reproductive stage and [SB] bamboo after senescent
stage.
In spite of a similar resolution, the QuickBird images have
advantages when compared to aerial photographs, even
considering only the visual interpretation. At the satellite image,
the tonal variations in the near-infrared band, allows the
discrimination of different phenologic phases of the bamboo
that are not evident in the aerial photographs, even if the
bamboo is at the early senescent stage. The temporal analysis
using QuickBird images is of promising importance in this case
to characterize the life cycles of these dominant bamboos
1291
'Mm