Non-Destructive Allometric Modeling for Tree Volume Estimation in Tropical Dry Deciduous Forests of India Using Terrestrial Laser Scanner

Spatial biomass estimation using remote sensing at a regional or national scale requires ground reference plots for developing calibration models and product validation. The reference plot above-ground biomass (AGB) estimates rely on allometric models, which estimate tree-level AGB through measurements of tree diameter, height, and species information. Recent developments in applying terrestrial laser scanning (TLS) to measure 3D canopy structure in great detail have enabled new opportunities for extracting tree volumes non-destructively. In this study, we aim to use 3D point clouds from TLS to model individual trees to extract tree volumes and thereby generate a framework for developing local allometric equations in tropical dry deciduous forests of Betul, Madhya Pradesh, India. We have used TLS scans of 127 individual trees across different species to generate site-level and species-specific allometric models. The generated TLS-based allometric models using diameter and height as predictor variables are validated with an independent destructive sampling dataset (n = 25 trees). Our TLS allometry models indicate superior predictions with lower error estimates (root mean square error (RMSE) of 10.9% (concordance correlation coefficient [CCC] = 0.98) compared to the equivalent traditionally used volume equations over the study site. We have also evaluated the uncertainty due to sample size and found that the prediction error stabilizes when the number of samples exceeds 100 trees. The results imply that TLS data can potentially increase the range and sampling size of allometric equations through non-destructive volume estimation to improve the traditional allometric models and reduce uncertainty in landscape-level biomass estimates.