Above-ground biomass estimation from LiDAR data using random forest algorithms
Leyre Torre‐Tojal, Aitor Bastarrika, A. Boyano, José Manuel López-Guede, Manuel Graña
Abstract
Random forest (RF) models were developed to estimate the biomass for the Pinus radiata species in a region of the Basque Autonomous Community where this species has high cover, using the National Forest Inventory, allometric equations and low-density discrete LiDAR data. This article explores the tuning for RF hyperparameters, obtaining two models with an R2 higher than 0.7 using 2-fold cross-validation. The models selected were applied in Orozko, a municipality with more than 5000 ha of this species, where the model predicts a biomass of 1.06–1.08 Mton, which is between 16–18 % higher than the biomass predicted by the Basque Government.
Topics & Concepts
Biomass (ecology)Random forestLidarTree allometryHyperparameterEnvironmental scienceAllometryAlgorithmPinus radiataEstimationForest inventoryStatisticsRemote sensingMathematicsForestryComputer scienceEcologyGeographyAgroforestryForest managementBiologyArtificial intelligenceEngineeringBiomass partitioningSystems engineeringRemote Sensing and LiDAR ApplicationsForest ecology and managementForest Ecology and Biodiversity Studies