International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
The available secondary data on natural resources and 
characteristics of coffee production systems of these regions 
was reviewed. To complement this information, field surveys 
were carried out in each area, in order to assess the relationships 
between coffee and the environment and collect georeferenced 
data from coffee fields, soils and geomorphology. Topographic 
maps from the Brazilian Institute of Geography and Statistics 
(IBGE), at the scale of 1:50.000 were used as à cartographic 
sis. Published soil surveys, geologic and geomorphic maps 
were also used, as well as aerial photographs, at the scale of 
1:25.000, and images from the TM Landsat 7 satellite, from 
1999 and 2000. The implementation of a digital database for the 
pilot arcas and the treatment of the satellite images were carried 
out using the geographic information system SPRING. 
Thematic layers of different physiographic aspects of the study 
regions, such as urban areas, roads and drainage networks, 
contours and spot heights information, were obtained from the 
topographic maps by digitizing. The areas occupied by coffee, 
surveyed and georeferenced during field surveys, were also 
digitized to create the correspondent digital overlay in the 
database. 
The TM Landsat images were treated in the SPRING's IMAGE 
module. Only three spectral bands were used for the 
classification of the image, viz. band 3 (red), band 4 (near 
infrared) and band 5 (mid infrared), since these bands represent 
more than 80% of the spectral information. Controlled samples 
of the main land cover types were used to train the classifier. 
These samples were obtained from areas identified and 
georeferenced during the field surveys. The segmentation of the 
images was performed using a region growing method and a 
supervised classification was carried out using the Maxver 
classifier (maximum likelihood algorithm available in the 
SPRING) on band 4. The thematic maps generated from the 
classified images identified the following land cover classes: 
Productive coffee: coffee fields with plants with more than 4 
years of age, 2 meters or more of height, and canopies covering 
more than 50% of the ground; Forest: correspondent to areas 
occupied by natural vegetation from “cerradäo” (Brazilian 
savanna woodlands) to remnants of semi deciduous Atlantic 
forest; Other uses: correspondent to areas occupied by small 
size natural vegetation (bush trees and other types of cerrado), 
natural and cultivated pastures, annual crops and even coffee 
fields still in formation, i.e. recently planted coffec areas which 
covered less than 50% of the ground; Bare Soil: 
comprehending areas prepared for cultivation and urban areas. 
These classes were chosen taking into account the main interest 
of the project, which was the coffee and the difficulty to 
distinguish this crop among natural forest and other land cover 
classes that spectrally overlap with it when sensed at the coarse 
spectral resolution of the TM Landsat. The thematic layer 
Actual Land Use was generated from the classified images. 
The Delaunay triangulation was used to construct a TIN and 
create a digital elevation model (DEM) and a digital terrain 
model (DTM) for each area, based on the digital contours and 
spot heights data obtained from the topographic maps. Slope, 
aspect and hypsometry class maps were derived from these 
models. The slope classes used in the segmentation of the DTM 
were selected based on the relations between geomorphology 
and the soils distribution, according to the field observations of 
the regional landscapes studied (Andrade ef al 1998). The 
geomorpho-pedologic model, validated for the regions of Sáo 
Sebastiào do Paraíso and Machado in field campaigns, is 
presented in Table 1. 
According to the Brazilian Soil Classification system, the main 
soil classes occurring in these regions are Latosols (soils with 
oxic B horizons), Argisols (soils with argillic B horizons), 
Cambisols (soils with cambic B horizons) Neosols (soils 
without diagnostic B horizons — Litholic and Alluvial soils) and 
Gleysols (Hidromorphic soils). As shown in table 1, Latosols 
occur predominantly where the slope gradient is lower than 
12%, although Hidromorphic and Alluvial soils might occur at 
flooded plains where the slope gradient is lower than 3%, 
Argisols occur where the slope gradient is higher than 12% ang 
lower than 45%, although Cambisols and Lithosols might occur 
locally where the slope gradient is higher than 24%. For the area 
of Sdo Sebstido do Paraiso the soil map was obtained from the 
overlay of the slope classes map and the geology domains map, 
according to the criteria established in Table 2. At the region of 
Machado however, the geology was more homogeneous and 
geomorphology and altitude were the primary factors 
determining the distribution of soil types over the local 
landscape. Therefore the model used the criteria shown in Table 
3. The soil map of the pilot area of Patrocínio was obtained 
from an unpublished pedological survey carried out by the soil 
research center of EMBRAPA. The mapping units delimited 
were checked in the field and adjusted to the scale of the work. 
The maps generated were crossed using the LEGAL/SPRING 
program. The thematic layers Land Use Classes x Soil Classes 
and Land Use Classes x Slope Classes, for cach pilot area, were 
overlayed and the quantitative relations were evaluated. 
3 RESULTS 
3.1 Study area of Patrocínio 
The results of the crossed tabulations (Land Use Classes x Soil 
Classes and Land Use Classes x Slope Classes) of the study 
area of Patrocínio are presented in tables 4 and 5, respectively. 
The geomorphology of the pilot area of Patrocinio is 
represented, predominantly, by flat to gently sloping surfaces of 
great extensions. In these areas Red Latosols and Yellow Red 
Latosols occur and are discriminated according to the content of 
iron of the parent material. Coffee is cropped in this 
environment, occupying 13, 86% of the lands of the pilot area, 
distributed mainly in the areas with plain and gently undulated 
relief. In this region the crop is characterized by large fields, 
generally in contiguous areas of great dimensions, where 
Latosols, correspondent mainly to the Ld! + Lea9 and Led16 + 
Ldl mapping units occur. Due to climactic conditions, irrigation 
is a frequent agricultural practice and mechanic harvest is also 
favored by the geomorphologic characteristics. These practices 
condition the management of the crop in the region and 
influence many of its parameters, such as maximum plant high 
(2.2 m) and planting density (3.8 m x 0.6 m). These parameters 
influence the percentage of soil cover by the coffee plants 
canopies, favoring the interpretation and survey of coffee fields 
in TM/Landsat satellite images, as well as propitiating the 
automatic classification and mapping of land use classes. 
3.2 Study area of Machado 
According to the geological maps available (DNPM/CPRM, 
1979), the pilot area of the region of Machado presents 
homogeneous geology, corresponding to the Complexo 
Varginha, constituted basically by ophthalmic gneiss and 
migmatites, whose mineral and chemical compositions do not 
Internation 
MA 
Slope Cla: 
0-129 
  
20-45% 
  
>45% 
Table 2. 
* Geologicc 
Qa: Quater 
TQi: Undiff 
Kb: Bauru 
coarse sand 
KJsg: Sdo b 
KJb: Sao Be 
locally with 
Pci: Tubare 
developmen, 
stratification 
** This soil 
  
Slope 
Class 
0-399 [— 
  
  
0-12% |— 
  
12-24% |— 
  
24-45% 
  
>45% |— 
et tis 
  
  
reflect in vat 
obtained fro 
the model es 
evaluation, p 
fieldwork, sc 
distributed i 
between slop 
these soil-m: 
described an 
Tables 6 an 
between Lan 
X Slope Clas: