MODELLING GROSS PRIMARY PRODUCTION OF TROPICAL FOREST BY REMOTE SENSING

The application of remote sensing has provided an opportunity to improve the estimation of gross primary production (GPP) on a regional scale. Several models to estimate GPP of homogeneous ecosystems, such as agricultural areas, entirely based on remote sensing data exist, but models to describe more heterogeneous areas are less common. Thus, the aim of the study was to evaluate the GPP estimated by different remote sensing methods in an Amazon-Cerrado transition forest in Mato Grosso, using MODIS spectral data. Two models, known as the temperature and greenness model (TG) and the vegetation index (VI) model, were used to estimate seasonal and interannual variations in GPP. Our results indicated that the TG and VI models were incapable of reproducing the seasonal variation in GPP, because the lack of correlation between vegetation indices and the GPP measured from tower-based eddy covariance (GPP EC ). Furthermore, the time series of the enhanced vegetation index (EVI) was delayed by 2 months with GPP EC . The results presented in this paper highlight some of the complexities in validating satellite products. Further study over a variety of Brazilian forests is needed to quantitatively assess the TG and VI and other methods to improve their accuracy.