In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Yol. XXXVIII, Part 7B 
536 
2. METHODOLOGY 
The study area includes the states in south central Brazil 
monitored by the Canasat project: Sao Paulo, Goiás, Mato 
Grosso, Mato Grosso do Sul, Minas Gerais and Paraná. 
Although south central Brazil includes ten states, the South- 
Central region referred to in this article includes only the six 
states of major sugarcane production that have an area of 2.63 
million km 2 , approximately 31% of the Brazilian territory. 
These states are responsible for more than 87% of sugarcane 
production in the country (UNICA, 2009). Figure 1 illustrates 
the location of the Brazilian states covered by the Canasat 
project. 
80’0'0"W 70°0'0"W 60"0'0"W 50”0'0"W 40°Q'0"W 
1 1 1— — 1 1 
Figure 1. States monitored by the Canasat project: Sao Paulo - 
SP, Goiás - GO, Mato Grosso - MT, Mato Grosso do Sul - MS, 
Minas Gerais - MG and Paraná - PR. 
From the first maps generated by the Canasat project (Rudorff 
et al., 2005), which were produced by means of digital and 
visual classification of the satellite images, annual updates were 
made through visual interpretation of a temporal sequence of 
images for each crop year. In developing the map for the current 
season, the map from the last season was used as a basis. Two 
steps were followed: visual interpretation of the images and 
reclassification of isolated pixels. Although digital classification 
was used only in the first mapping, it is less relevant in updating 
maps for the current season because this procedure tends to 
commit many errors of omission and inclusion, making visual 
classification and editing necessary to obtain a satisfactory 
thematic map. 
In the visual interpretation step, the matrix edit function within 
the SPRING program was used (Camara et al., 1996). This 
function permits modification or correction of the classification 
results and definition of new class polygons. In this procedure, 
the interpreter used images from several dates contained in the 
database to confirm the appropriate class for each sugarcane 
stage. To homogenize the results, the resulting map from each 
database was revised by a more experienced interpreter and 
subsequently subjected to reclassification of isolated pixels. 
This technique enables the controlled and automatic correction 
of isolated pixels or groups of pixels that were incorrectly 
classified or unintentionally inserted by the interpreter. 
To distinguish sugarcane from other targets, images from 
specific periods of the plant development were used, that is, 
images from dates when the sugarcane crop was well developed 
(January to March). Therefore, the mapping of areas to be 
harvested in the current season is based on images acquired 
prior to the beginning of the harvest season in April. However, 
it is more likely to acquire cloud free images from April on and, 
therefore, whenever necessary images from April and May were 
also used in the mapping procedure, especially during the 
revision part. 
Mapping and estimation of sugarcane cultivated area available 
for harvest incorporated three distinct classes: 
Areas of sugarcane cultivation were identified in images from 
the TM/Landsat-5 sensor. As an alternative to the TM images 
that were obscured by cloud cover, images from the CCD 
(Charge Couple Device) sensor on board of the CBERS-2 and 
CBERS-2B satellites were used (Epiphanio et al., 2007). For 
each orbit/point of the TM/Landsat sensor, a database of images 
obtained by the two sensors on the dates of interest was created. 
All images were registered using first-order polynomials and 
nearest-neighbor interpolations based on orthorectified mosaics 
of TM/Landsat-7 images obtained from NASA (NASA,2007). 
Sugarcane available for harvest - This class included areas of 
sugarcane that could be harvested in the current season. This 
class included three groups: 
a) Ratoon Cane - This term refers to areas of sugarcane 
available for harvest from sprouting after first or more cuts; 
1 a) Apr. 24 th , 2007 1b) Jun. 11 s ', 2007 1 c) Mar. 25 m , 2008 1d) Sep. 17», 2008 
2a) Sep. 15», 2007 2b) Mar. 25 th , 2008 2c) Apr. 26 th , 2008 2d) Dec. 06"’, 2008 
Figure 2. Temporal sequence of TM/Lansat-5 images [color composition 4(R)5(G)3(B)] illustrating the identification of areas of 
expansion (2.1) and under renovation (2.2).