Aboveground Biomass Modeling from Field and LiDAR Data in Brazilian Amazon Tropical Rain Forest

Abstract:

Tropical forests are an important component of global carbon stocks, but tropical forest responses to climate change are not sufficiently studied or understood. Among remote sensing technologies, airborne LiDAR (Light Detection and Ranging) may be best suited for quantifying tropical forest carbon stocks. Our objective was to estimate aboveground biomass (AGB) using airborne LiDAR and field plot data in Brazilian tropical rain forest. Forest attributes such as tree density, diameter at breast height, and heights were measured at a combination of square plots and linear transects (n=82) distributed across six different geographic zones in the Amazon. Using previously published allometric equations, tree AGB was computed and then summed to calculate total AGB at each sample plot. LiDAR-derived canopy structure metrics were also computed at each sample plot, and random forest regression modelling was applied to predict AGB from selected LiDAR metrics. The LiDAR-derived AGB model was assessed using the random forest explained variation, adjusted coefficient of determination (Adj. R²), root mean square error (RMSE, both absolute and relative) and BIAS (both absolute and relative). Our findings showed that the 99th percentile of height and height skewness were the best LiDAR metrics for AGB prediction. The AGB model using these two best predictors explained 59.59% of AGB variation, with an Adj. R² of 0.92, RMSE of 33.37 Mg/ha (20.28%), and bias of -0.69 (-0.42%). This study showed that LiDAR canopy structure metrics can be used to predict AGC stocks in Tropical Forest with acceptable precision and accuracy. Therefore, we conclude that there is good potential to monitor carbon sequestration in Brazilian Tropical Rain Forest using airborne LiDAR data, large field plots, and the random forest algorithm.