HIGH-RESOLUTION IMAGE TO MAP BAMBOO-DOMINATED GAPS IN THE 
ATLANTIC RAIN FOREST, BRAZIL 
L. S. Araujo 3 , G. Sparovek 3 , J. R. dos Santos b , R. R. Rodrigues 3 
a ESALQ, University of Sâo Paulo, Piracicaba, Sâo Paulo, Brazil - (lsaraujo, gerd, rrr)@esalq.usp.br 
b DSR, National Institute for Space Research, Sâo José dos Campos, Sâo Paulo, Brazil-jroberto@ltid.inpe.br 
KEY WORDS: Ecology, Land Cover, Forestry, Landscape, Interpretation, Quickbird, High Resolution 
ABSTRACT: 
This study presents the results referring to the analysis of QuickBird high resolution satellite images to establish patterns associated 
to the landscape dynamics in bamboo forests at the Intervales Park (Sao Paulo State, Brazil), considering its ecologic aspects. The 
methodology was based on visual image interpretation, as a preliminary step for digital processing. The results show the possibilities 
for the use of these high resolution data to map forests with bamboo dominance as well as for monitoring phenologic cycles of these 
dominant species. The visual interpretation was an adequate technique for our study, emphasizing that the knowledge of the 
ecological aspects from these forest typologies was of fundamental importance to understand the patterns mapped with this image. 
1. INTRODUCTION 
Presently remote sensing data are fundamental tools to survey 
natural resources, being of fundamental importance for decision 
taking in several areas of public policies. In Brazil, several 
methodologies using orbital data were developed and applied 
exclusively to map and monitor vegetation including several 
Brazilian biomes such as presented at the “Atlas of forest 
remnants from the Mata Atlantica" (SOS Mata Atlantica, INPE 
2002). 
At Atlantic Forest (Mata Atlantica), composed presently by 
fragments distributed along the Brazilian coast, the region 
known as “Continuous of Paranapiacaba” is considered one of 
the most important remnants of this biome. Located at southern 
Sao Paulo State and formed by the junction of Parks Intervales, 
Turlstico do Alto Ribeira and Carlos Botelho and the Ecologic 
Station of Xitue, the Continuous includes approximately 150 
thousand ha of vegetation composed almost completely of 
Ombrophilous Dense Forest (BIOTA/FAPESP, 2008). Recent 
diagnostics in the field however, associated with the analysis of 
aerial photographs, present extensive forest areas with 
dominance of bamboos (Sao Paulo, 2007), which were not 
detected on previous surveys (BIOTA/FAPESP, 2008; SOS, 
INPE, 2002). This fact evidenced the need to map the 
distribution and monitor the possible expansion of these forest 
formations at a compatible scale with this landscape. 
The bamboos belong to the family of the Poaceae and are 
typical components of the humid forest from Mata Atlantica 
(Rizzini, 1979). Due to its’ characteristics of fast colonization 
from the rhizomes and of its’ invasive habits, they are 
frequently related to disturbed environments (Wong, 1991), to 
clearings of natural and human origin, interfering on both forest 
structure and regeneration (Griscom & Ashton, 2003). Besides 
that, they have a long life cycle between 20 and 60 years, 
culminating at a phase of sexual reproduction with a high 
production of seeds and massive mortality of population 
(Janzen, 1976), restricting the knowledge of the life cycle from 
these plants to a few species. These associated propagation 
characteristics allow its dominant establishment which results in 
differentiated landscape patterns, of fundamental importance to 
understand the ecological aspects associated to these bamboo 
sites in studies of forest dynamics. 
In Brazil studies of bamboo formations with remote sensing 
techniques were performed in the SW Amazon, where there are 
extensive areas dominated by the genus Guadua (Nelson et al., 
2006). Using a series data from 28 years of Landsat-MSS, 
Landsat-TM and MODIS/TERRA, Nelson & Bianchini (2005) 
monitored the life cycle of the bamboo population of genus 
Guadua in Acre State, during two population mortality events 
in 1976 and 2003-04, learning on the reproductive cycles of 
these species. Despite differences in structure and floristic 
composition, forest classes with and without bamboo were not 
well discriminated with JERS-1 (Santos et al., 2000) and 
ASTER/Terra (Mendoza et al., 2004) image data in sections of 
the Amazon. 
In Acre State, bamboo forests occupy large areas, with one or 
two dominant species. At Intervales Park, in the Continuous of 
Paranapiacaba, field observations show clearings occupied by 
different bamboo species occurring jointly with sections of 
dense forest. This landscape mosaic and the extremely 
mountainous topography of the region, make mapping of these 
areas difficult with medium resolution data. Being so, high 
resolution images become an additional information source for 
these studies on bamboo, with similar resolution of aerial 
photographs, but with distinct spectral characteristics which 
allow detailed mapping of forest features using automatic 
classification procedures. 
Taking into account the complexity of these high resolution 
data and consequently of the associated processing, the visual 
interpretation is still a quite frequently used procedure, as an 
alternative or in parallel to digital procedures (Buchanan et al., 
2008; Tottrup et al., 2007). Based on parameters such as color, 
texture, shape and pattern (Lillesand & Kiefer, 1994), this 
technique allows the direct inclusion of knowledge from the 
analyst in the field during the mapping process, fundamental to 
understand this forest dynamics in areas with bamboos. 
In this context, the objective of this paper is to evaluate the 
potential of high-resolution images to establish patterns