Modelling above-ground biomass of Eucalyptus bosistoana F.Muell. and Eucalyptus globoidea Blakely

Abstract

Background: Estimation of forest biomass has become critical as afforestation has been proposed to sequester carbon from the atmosphere in order to mitigate climate change. New Zealand Dryland Forestry Innovation (NZDFI), in collaboration with the University of Canterbury’s School of Forestry and the Marlborough Research Centre, has initiated a research and development programme to gather seed, breed, propagate, identify site limitations, model growth, investigate silviculture, and develop wood products from a suite of eucalypts that grow durable heartwood. The aim is to supply naturally durable wood for uses that formerly required either imports of durable wood or copper-chrome-arsenate treated pine.

Methods: As part of a project examining land-use and greenhouse gas budget case studies in Marlborough, New Zealand, we collected and summarised data describing above-ground biomass (AGB) of Eucalyptus bosistoana F.Meull., and Eucalyptus globoidea Blakely trees across a wide range of combinations of height (h) and diameter at breast height (dbh). One hundred and eleven trees were felled, separated into stems, branches and foliage, and the components were weighed in the field. Subsamples of these tree parts were collected and weighed in the field after separating bark from stem discs. The subsamples were dried in an oven at 105°C, and then weighed. Ratios of dry to wet weights for samples were applied to total green weights from the field in order to calculate AGBs of tree components. Systems of non-linear equations were simultaneously fitted to the data to ensure additivity; that sums of estimates of tree part AGBs versus dhb, h and slenderness (h/dbh) equalled estimates from a model of total tree AGB versus the same independent variables. The study also included the development of a plot-level estimation model of above-ground CO2-e/ha for E. globoidea and its incorporation in an on-line growth and yield simulator. Moreover, a comparison of two pathways to estimating AGB by aerial LiDAR was made: One including estimates of dbh and h from LiDAR and applying the tree-level equations developed in this study, and one going directly from LiDAR metrics to estimates of AGB.

Results: A system of models created for both species with a dummy variable denoting species yielded the least biased residuals, with 22 coefficients estimated in one simultaneous fit. Standard errors varied with plant part and with the size of the prediction, requiring transformations prior to fitting. R² values also varied with part, but were typically between 0.96 and 0.98. An exception was foliage and seeds which were influenced by one tree with an unusually high loading of seeds. The standard error for plot level estimates of CO2-e was 1.9 tonnes CO2-e /ha and residuals were relatively unbiased. Directly predicting individual tree AGB from LiDAR metrics yielded less biased estimates than predicting dbh and h and then using those estimates to predict AGB.

Conclusions: A system of related, additive equations with a dummy variable denoting species represented the above-ground biomass of Eucalyptus globoidea and Eucalyptus bosistoana with precision adequate for prediction of biomass for fuel and carbon storage to mitigate climate change. Direct predictions of biomass from LiDAR metrics were less biased than predictions of biomass from tree height and diameter at breast height that were in turn predicted from LiDAR metrics.