The main objective was to study the multitemporal expansion of agriculture for 33 years using three different satellites/sensors, by applying Principal Components Analysis techniques in order to generate the components of brightness and greenness for each dataset. The use of these components for the Change Vector Analysis can thus provide information on the intensity and type of change occurred. We used MSS/Landsat, TM/Landsat and CCD/CBERS, acquired between 1975 and 2008. The Kappa coefficients ranged from 0.18 to 0.41, indicating that the change of Vector Analysis had slight or fair agreement with visual analysis. Assuming a significance level of 0.05, it was verified that the result of analysis by Change Vector Analysis is better than a random classification. In general, the errors are due to spectral confusion associated with natural or anthropogenic land use, such as “natural grassland� and grazing, and increases in plant biomass, which may refer to forest regeneration or development of agricultural crops. Change Vector Analysis was useful for detecting changes and it accepts the use of different parameters and considers their variation over time. As input data, the principal components are direct and rapid means for generating information of brightness and greenness of a particular scene. The principal components are feasible in studies involving the analysis of the variation of these parameters.
multisensors; change detection; principal component analysis; change vector analysis; agriculture
