Modelos de estimativa de biomassa aérea utilizando dados RapidEye para a Floresta Nacional do Tapajós-PA

The objective of this study was to develop statistical models for estimating aboveground biomass (AGB) at Tapajos National Forest, Pará, Brazil, using spectral metrics derived from RapidEye. Measurements of diameter at breast height (DBH) and tree height were collected for 88 forest inventory plots (50 m x 50 m). All trees were identified to the genus and/or species level and their biomasses were estimated using allometric equations. The explanatory variables were extracted from the five spectral bands of the RapidEye satellite (5 m spatial resolution) and included individual bands, band ratios, and vegetation indices. Biomass estimation models were fitted using multiple linear regression and the non-parametric algorithm Random Forest. The predictive performance of the models was assessed based on the coefficient of determination (r2) and the root mean square error (RMSE) calculated using a cross-validation procedure. The best regression model selected included three variables and presented a cross-validation r2 of 0.67 and a RMSE of 95,4 Mg ha-1 (50%). The Random Forest algorithm presented a better performance, with an r2 of 0.75 and a RMSE of 84,1 Mg ha-1 (45%). We conclude that metrics derived from the RapidEye sensor have the potential to explain a large portion of the variability in biomass at Tapajos, when combined with a more powerful statistical framework such as Random Forest.