Using terrestrial laser scanning data to estimate large tropical trees biomass and calibrate allometric models: A comparison with traditional destructive approach

Abstract

Calibration of local, regional or global allometric equations to estimate biomass at the tree level constitutes a significant burden on projects aiming at reducing Carbon emissions from forest degradation and deforestation. The objective of this contribution is to assess the precision and accuracy of Terrestrial Laser Scanning (TLS) for estimating volumes and above‐ground biomass (AGB) of the woody parts of tropical trees, and for the calibration of allometric models. We used a destructive dataset of 61 trees, with diameters and AGB of up to 186.6 cm and 60 Mg respectively, which were scanned, felled and weighed in the semi‐deciduous forests of eastern Cameroon. We present an operational approach based on available software allowing the retrieving of TLS volume with low bias and high accuracy for large tropical trees. Edition of the obtained models proved necessary, mainly to account for the complexity of buttressed parts of tree trunks, which were separately modelled through a meshing approach, and to bring a few corrections in the topology and geometry of branches, thanks to the amapstudio‐scan software. Over the entire dataset, TLS‐derived volumes proved highly reliable for branches larger than 5 cm in diameter. The volumes of the remaining woody parts estimated for stumps, stems and crowns as well as for the whole tree proved very accurate (RMSE below 2.81% and R² above of .98) and unbiased. Once converted into AGB using mean local‐specific wood density values, TLS estimates allowed calibrating a biomass allometric model with coefficients statistically undistinguishable from those of a model based on destructive data. The Unedited Quantitative Structure Model (QSM) however leads to systematic overestimations of woody volumes and subsequently to significantly different allometric parameters. We can therefore conclude that a non‐destructive TLS approach can now be used as an operational alternative to traditional destructive sampling to build the allometric equations, although attention must be paid to the quality of QSM model adjustments to avoid systematic bias.

LeWoS: A universal leaf‐wood classification method to facilitate the 3D modelling of large tropical trees using terrestrial LiDAR

“…Sixty‐one large tropical trees from the Eastern of Cameroon were included in this study. This set of trees was previously used for estimating large tropical tree biomass from TLS point clouds and calibrating allometric models (Momo Takoudjou et al, , ). In total, the dataset covered 15 different species.…”

“…The acquired high‐density point clouds allow very precise measurements of tree structures in a non‐destructive way. These measurements include forest inventory (Liang et al, ) leaf angle distribution (Vicari, Pisek, & Disney, ), structural parameters (Trochta, Krůček, Vrška, & Král, ; Wang, Hollaus, Puttonen, & Pfeifer, ), above‐ground volume and biomass (AGB) (Calders et al, ; Gonzalez de Tanago et al, ; Momo Takoudjou et al, ). Moreover, allometric equations can be non‐destructively performed with data derived from TLS (Lau et al, ; Momo Takoudjou et al, ).…”

“…For quantitative ecology, tree volume or AGB is the key feature of interest (Disney et al, ). Recent studies have shown that TLS, combined with QSM reconstruction methods, can be used to accurately estimate above‐ground volume and AGB for large tropical trees (Gonzalez de Tanago et al, ; Momo Takoudjou et al, ). The accuracy of volume and AGB estimates depends on the quality of the reconstructed QSM, which in turn, relies on a well‐filtered point cloud of tree woody components.…”

“…The accuracy of volume and AGB estimates depends on the quality of the reconstructed QSM, which in turn, relies on a well‐filtered point cloud of tree woody components. In general, tropical scenes are more challenging to reconstruct because of a poorer point cloud quality in the tree crown (Momo Takoudjou et al, ). Also, laborious and tedious manual removal of leaf points has to be carried out, which can require up to 15 hr per tree (Momo Takoudjou et al, ).…”

“…In general, tropical scenes are more challenging to reconstruct because of a poorer point cloud quality in the tree crown (Momo Takoudjou et al, ). Also, laborious and tedious manual removal of leaf points has to be carried out, which can require up to 15 hr per tree (Momo Takoudjou et al, ). An automated processing chain would greatly facilitate the usage of TLS for quantifying 3D structures of large tropical trees.…”