Aboveground Woody Biomass Estimation of the Brazilian Cerrado Biome Using Data Integration

Abstract

The Brazilian Cerrado consists of a highly heterogeneous tropical savanna, and its biomass stocks are highly variable. Mapping and monitoring these stocks is not trivial. To address this challenge, we built an aboveground woody biomass (AGWB) model for the Cerrado biome using 30-m resolution optical satellite imagery (Landsat-5 and Landsat-8), 25-m resolution SAR imagery (ALOS and ALOS-2), and a set of plot-based and LiDAR-derived AGWB estimates. We implemented both a Classification and Regression Tree (CART) and a Random Forest (RF) algorithm to model AGWB over the native vegetation for the year 2019 in the Cerrado. The RF algorithm resulted in a slightly better result $(\\mathrm{R}^{2}=53\\%;\\text{ rel. RMSE} =57\\%)$ than, the CART model $(\\mathbf{R}^{2}=45\\%\\ \\text{ rel. RMSE} =63\\% 1$, but our map shows an underestimation of very high AGWB (negative bias over 200 t.ha-1) and a slight overestimation of low AGWB (positive bias).