Effectiveness of BFAST algorithm to characterize time series of dense forest, agriculture and pasture in the Amazon region

Abstract

Vegetation is one of the most important components of ecosystems, attracting attention and interest of the scientific community due to its undergoing constant transformation. The remote sensing systems provide data to detect, identify, map and monitor these changes. This study aimed at (1) evaluating the effectiveness of the BFAST algorithm to characterize time series of dense forest, agriculture and pasture in the Amazon region; (2) performing statistical tests in order to compare these series, and (3) fitting models to predict future values. By using the cumulative sums test, the time series of the three classes of land use were statistically different from each other, when comparing in pairs. As the series were different, the time series analysis of remote sensing data was useful in the identification and classification of different types of land use. The use of adjusted models to predict future values of the time series has proven effective for the use of Agriculture and Pasture, but not for the Forest class. It is concluded that the BFAST algorithm characterization of time series for the subsequent adjustment of models was useful for predicting harvests, considering the Agriculture use class.