Estimating carbon biomass in forests using incomplete data

Lahiru Suranga Wijedasa, Anuj Jain, Alan D. Ziegler, Theodore Alfred Evans, Tak Fung

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)


Historical vegetation studies have been of limited use in total aboveground biomass (AGB) estimation because they only report incomplete data consisting of tree diameter-class distributions or plot-level summaries, rather than data on each tree individual. To address this issue, we assessed an existing method (ST-n) and developed three new methods (ST-p, PL-n and PL-p) for estimating total AGB using only incomplete data from tropical forest plots. ST-n and ST-p apply to studies with tree diameter-class distributions (or stand tables, “ST”), whereas PL-n and PL-p apply to studies with plot-level summary variables (“PL”) in the form of total tree basal area, mean tree diameter, and total number of trees. ST-n and PL-n are non-parametric (“n”) methods that do not impose any form on the underlying distribution of tree diameters. In contrast, ST-p and PL-p are parametric (“p”) methods that involve fitting probability distributions of tree diameter to the data. We applied the methods to incomplete data from 58 1-ha plots in Panama and 300 1-ha pseudo-plots (generated by randomly sampling tree diameters from empirical distributions for three larger plots) in Southeast Asia, and four allometric equations. For these two regions and equations, ST-p gave low total proportional errors (TPEs, as measured by proportional root-mean-square error) of 1%–8%. In contrast, ST-n gave moderate to large TPEs of 10%–66%. PL-n and PL-p gave low to moderate TPEs of 5%–30%. The methods have great potential to expand the pool of large-scale baseline AGB assessments to historical studies with incomplete data.

Original languageEnglish
Pages (from-to)397-408
Number of pages12
Issue number2
Early online date18 Nov 2020
Publication statusPublished - Mar 2021


Dive into the research topics of 'Estimating carbon biomass in forests using incomplete data'. Together they form a unique fingerprint.

Cite this