Estimation of Above-Ground Biomass for Dendrocalamus Giganteus Utilizing Spaceborne LiDAR GEDI Data

Abstract

Bamboo forests are paramount in forest ecosystems due to their rapid growth and huge carbon sequestration potential. Consequently, the estimation of bamboo biomass emerges as a critical research endeavor in forest remote sensing. The study used global ecosystem dynamics investigation (GEDI) as the data source to estimate the regional-scale above-ground biomass (AGB) of Dendrocalamus giganteus. The outcomes reveal that 1) the results showed that the power function emerged as the most efficacious model, with coefficient of determination (R2) = 0.87 and root mean square error (RMSE) = 0.00051 Mg, in estimating the AGB of Dendrocalamus giganteus. 2) Based on the feature importance ranking of Random Forest, five variables were selected from the 40 extracted from GEDI, achieving RMSE = 8.21 Mg/ha and mean absolute error (MAE) = 6.12 Mg/ha. The optimal variational function model, obtained using GS+9.0 software, revealed an exponential model for pgap_theta_a6, pai, and rg_a5, and a spherical model for leaf_on_doy and pft_clas. 3) The stacking model with K-nearest neighbors as the metalearner (R2 = 0.93, RMSE = 7.04 Mg/ha, MAE = 4.44 Mg/ha) outperformed the individual random forest regression and support vector regression.