Pantropical and continental stand-level biomass models

Background : This study utilized a pantropical tree database to simulate forest ground plots across the tropical zone. The aims were to (i) generate pantropical and continental stand-level aboveground biomass (AGB) models, (ii) compare performances of stand- and tree-level models in AGB prediction per unit area, and (iii) quantify gains in accuracy and precision by adding Lorey’s height as a model predictor. Forest stand variables such as basal area and tree density were calculated and then used as predictors in the AGB models. An additional predictor, Lorey’s height, was included and its contribution to the model performance was quantified. We estimated models at two scales: pantropical and continental [Tropical (T) Africa, T. Asia, T. Central America and T. South America]. Our models were compared with respect to accuracy and precision of predictions to the pantropical biomass model – referred to as ‘Model0’. Results: Our continental models reduced the mean error of AGB prediction from 51.9 to -0.1 Mg ha ^-1 in T. Central America, from -16.8 to -0.7 in Mg ha ^-1 T. Asia, and from 11.0 to -0.3 in Mg ha ^-1 in T. South America, relative to ‘Model0’. Globally, our pantropical model predicted AGB per unit area 3.7 times more accurately than ‘Model0’, but a bit less precisely. Most models improved in performance when adding Lorey’s height as a predictor. Conclusions: We recommend our models to predict forest AGB at large scales, highlighting that our models independent of Lorey’s height are an option that ensures accurate AGB predictions.