The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
1288
associated to the dynamics of landscape in bamboo forests at
Intervales Park, considering the ecologic aspects of this forest
physiognomy. The methodology was based on the visual
interpretation of images, as a preliminary step for the
application of digital processing. The results obtained integrate
a databank of a project for the analysis of the spatial and
temporal distribution of bamboo forests from Intervales,
subsidizing the discussion on the monitoring of such areas.
2. STUDY AREA
This study was performed in the Intervales Park (Figure 1), a
conservation unit of the Continuous of Paranapiacaba region,
located at southern Sao Paulo State, between geographic
coordinates S 24° 12’ - 24° 32’ and W 48° 03’ - 48° 32’. The
section selected for analysis, with an area of 2,300 ha is located
between the administrative office from this Park and the control
bases Carmo and Capinzal, that are characterized by a history
of intense natural and human disturbances during the time
before the constitution of this conservation reserve.
s 24“! 9' / o 48”29'
Figure 1. Localization of Intervales Park (PEI) in Sao Paulo
State/Brazil and the area under study within PEI (in yellow) at
Quickbird image (R4G2B1).
3. METHODOLOGY
This study was done using a QuickBird image, radiometric
resolution 11 bits (2048 gray levels), type Standard 2A, with
five spectral bands, namely 1 panchromatic, spatial resolution
of 0,6 m and 4 multi-spectral (blue, green, red and infrared),
spatial resolution of 2,4 m. The data-take was May 19th 2006,
with an area of 8,000 ha, including part of the park and a
section of its surroundings. A sample area was selected based
on more detailed knowledge in the field survey.
The mapping activity with the QuickBird image was supported
by a databank with historical records from the area (1962-2000)
and a set of field survey data with the description of dominant
landscapes, referring to four years of monitoring (2004-2008).
Taking into account that the use of high-resolution images
aimed to generate data to be integrated into a databank of the
project, an ortho-rectification of the image was not done.
Nevertheless the image was registered, considering as a base a
mosaic of aerial photographs from 2000 (scale 1:35.000).
Afterwards an image fusion procedure was undertaken, using
the principal components method (Pinho, Renno & Kux, 2005),
and the result was a fused image with a resolution of 0,6 m.
Image processing at the QuickBird scene was done using the
software package ENVI 4.1.
Based on information from the databank, an interpretation key
was established, considering different patterns found in the field
as a function of the QuickBird image bands that are best suited
for this mapping. The definition of mapping classes was based
on parameters of visual interpretation like texture and color,
associated to structural vegetation information such as height
and canopy density, besides information on the dominance of
bamboo. This information directed the visual interpretation of
the image, done in ArcView 3.2, at an approximate scale of
1:5,000, and the generation of a land cover map. Additionally
the thematic mapping classes were discussed as a function of
spectral responses of targets of interest.
4. RESULTS
Taking into account the different patterns found in the field, the
key for the interpretation of the QuickBird image (Table 2) was
elaborated with 9 classes of land cover: (1) water bodies, (2)
bare soil, (3) grass, (4) bush, (5) secondary succession with
dominance of bamboo in vegetative or reproductive stage, (6)
secondary succession with senescent bamboos, (7) secondary
succession with spaced bamboo sections, (8) advanced
secondary succession and (9) primary forest. These classes were
defined considering bands R4G2B1 of the fused image, an
image composition which was most adequate for this mapping
effort because it allows the distinction of different forest
categories. The result of the image interpretation was a thematic
map of land cover from a section of the PEI (Figure 3),
including the quantification of the respective areas by class
(Table 4), and by the analysis of spectral response of targets
(Figure 5).