New PhytologistPlant BiologyPeer-reviewed journal article
Beyond species means - the intraspecific contribution to global wood density variation
Fabian Jörg FischeriD, Jérôme ChaveiD, Amy E. ZanneiD, Tommaso JuckeriD, Alex FajardoiD, Adeline FayolleiD, Renato Augusto Ferreira de LimaiD, Ghislain VieilledentiD, Hans BeeckmaniD, Wannes HubauiD, Tom De MiliD, Daniel Wallenus, Ana María AldanaiD, Esteban Alvarez-Dávila, Luciana F. AlvesiD, Deborah M. G. ApgauaiD, Fátima Aparecida ArcanjoiD, Jean‐François BastiniD, Andrii BilousiD, Philippe BirnbaumiD, Volodymyr BlyshchykiD, Joli Rumi BorahiD, Vanessa BoukiliiD, J. Julio CamareroiD, Luisa CasasiD, , David Y. P. TngiD
Abstract
Wood density is central for estimating vegetation carbon storage and a plant functional trait of great ecological and evolutionary importance. However, the global extent of wood density variation is unclear, especially at the intraspecific level. We assembled the most comprehensive wood density collection to date (GWDD v.2), including 109,626 records from 16,829 plant species across woody life forms and biomes. Using the GWDD v.2, we explored the sources of variation in wood density within individuals, within species, and across environmental gradients. Intraspecific variation accounted for up to 15% of overall wood density variation (sd = 0.068 g cm-3). Sapwood densities varied 50% less than heartwood densities, and branchwood densities varied 30% less than trunkwood densities. Individuals in extreme environments (dry, hot, acidic soils) had higher wood density than conspecifics elsewhere (+0.02 g cm-3, [~]4% of the mean). Intraspecific environmental effects strongly tracked interspecific patterns (r = 0.83) but were only 20-30% as large and varied considerably among taxa. Individual plant wood density was difficult to predict (RMSE > 0.08 g cm-3; single-measurement R2 = 0.59). We recommend (i) systematic within-species sampling for local applications, and (ii) expanded taxonomic coverage combined with integrative models for robust estimates across ecological scales.
- organism: plant (other)
- organism: multiple species
- research focus: database / resource
- research focus: ecology & biodiversity
- research focus: development & morphogenesis
- research focus: chromatin & epigenetic regulation
- approach: Bayesian inference
- approach: mixed model
- approach: statistical model
In-depth summary
Claude Opus 5.5 · xhigh effort · 5 Nov 2026 · from the full text of version 1 (28 Aug 2025) · CC BY 4.0
Across 109,626 wood density records from 16,829 plant species (GWDD v.2), intraspecific variation explains 15% of global variance, tracks interspecific environmental gradients at 20-30% magnitude, and limits single-plant predictability.
Wood density is a foundational functional trait for terrestrial carbon stock estimation and ecological modeling, yet prior global compilations (including GWDD v.1) lacked metadata on plant organs, tissue types, and individual-level sampling needed to separate intraspecific variation from measurement error. Assembling GWDD v.2 with 6.7-fold more records and detailed sampling attributes enables global partitioning of within-individual, among-individual, and environmental drivers of wood density.
Key results
ShownIntraspecific variation and residual error account for 15% of global wood density variance, while taxonomic family, genus, and species differences explain 77%.
Across GWDD v.2 (109,626 wood density records from 16,829 resolved species in 166 countries; global mean = 0.56 g cm^-3, sd = 0.178 g cm^-3), variance was partitioned into family (30%), genus (30%), and species (17%) levels (combined 77%, sd = 0.156 g cm^-3), study methodology (8%, sd = 0.049 g cm^-3), and intraspecific/residual variation (15%, sd = 0.068 g cm^-3). · Abstract; The Global Wood Density Database; The extent and shape of intraspecific variation…; But intraspecific variation in wood density is…
ShownWithin individual plants, wood density varies substantially less in sapwood than heartwood and less in branchwood than trunkwood, with trunk-branch differences largely explained by sapwood fractions.
Sapwood variance was 50% lower than heartwood (sd = 0.144 vs. 0.197 g cm^-3; Pearson's r = 0.78, nspecies = 150) and branchwood variance was ~30% lower than trunkwood (0.120 vs. 0.145 g cm^-3; r = 0.67, nspecies = 2,018, mean decrease -0.023 g cm^-3), but restricting trunk-branch comparisons to sapwood reduced the variance difference to <20% (0.112 vs. 0.124 g cm^-3, r = 0.76, nspecies = 523). · Abstract; Estimating variation in wood density within…; Wood density variation within individuals; Intraspecific variation in wood density matters
ShownIntraspecific wood density shifts across climatic and edaphic gradients closely mirror interspecific patterns in direction but are 70-80% smaller in magnitude and vary widely across species.
Intraspecific environmental effects correlated strongly with interspecific effects (r = 0.83), increasing with standardized annual mean temperature (+0.012 g cm^-3) and water deficit (+0.010 g cm^-3 intraspecific vs. 0.038 g cm^-3 interspecific, species slope sd = 0.038 g cm^-3) and decreasing weakly with wind speed and soil pH (-0.003 g cm^-3 each). · Abstract; Wood density variation among individuals; But intraspecific variation in wood density is…
ShownTaxonomic hierarchical models substantially improve species-mean wood density estimates for undersampled taxa, but individual tree wood density remains difficult to predict from local conspecifics.
A single species measurement predicted the species mean no better than an unsampled genus average (RMSE = 0.084 vs 0.083 g cm^-3 across 1,667 species), whereas two measurements in a hierarchical model lowered species-mean RMSE to 0.038 g cm^-3; predicting an individual plant from one local conspecific yielded RMSE = 0.107 g cm^-3 (R^2 = 0.59, or RMSE = 0.086 g cm^-3 and R^2 = 0.71 with hierarchical pooling). · Abstract; The effect of intraspecific variation on wood…; The effect of intraspecific variation on wood…
Methods, data and reproducibility
Compiled and standardized 109,626 wood density records (converting airdry/ovendry values to basic wood density in g cm^-3 and resolving taxonomy via WorldFlora) across 16,829 species in GWDD v.2. Fitted Bayesian (brms/STAN) and ML (lme4) hierarchical linear mixed models and major axis regressions (lmodel2) to partition taxonomic, methodological, organ, and environmental variance and test prediction accuracy.
- Global Wood Density Database v.2 (GWDD v.2) wood and bark density records · woody plants (angiosperms, gymnosperms, tree-like monocots) · n = 109,626 wood density records from 16,829 species, plus 15,093 bark records · 10.5281/zenodo.16919509
- Archived analysis datasets and hierarchical modeling code · woody plants · n = 27 statistical models across filtered GWDD v.2 subsets · 10.5281/zenodo.16928343
- Global bioclimatic and soil predictor rasters (CHELSA/BIOCLIM+ 1981-2010, TerraClimate 1981-2010, SoilGrids) · n = 1 km and 5 km global spatial grids
- Methods & tools
- WorldFlora, World Flora Online (WFO), GlobalTreeSearch 1.7, R, brms, STAN, lme4, lmodel2, CHELSA/BIOCLIM+, TerraClimate, soilgrids
- Data & accessions
- https://www.naturalearthdata.com/
- Reagents & environment
- R environment with packages WorldFlora (World Flora Online June 2023), brms, STAN, lme4 ('bobyqa' optimizer), and lmodel2
- Reproducibility gaps
- The GWDD v.2 database (doi: 10.5281/zenodo.16919509) and analysis code archive (doi: 10.5281/zenodo.16928343) on Zenodo are embargoed until acceptance for publication.
Limitations (3)
- Within-individual comparisons relied on coarse binary categories (heartwood vs. sapwood, trunkwood vs. branchwood) that do not capture non-linear radial or longitudinal wood density gradients. · stated by the authors
- The database lacks local micro-environmental growth conditions and competitive neighborhood data (such as light availability, rockiness, or waterlogging) that shape individual-plant wood density. · stated by the authors
- Geographic coverage of known tree species is below 50% in many species-rich tropical regions, and most West African records lack precise geolocation coordinates. · stated by the authors
Read this if you estimate forest carbon stocks, parameterize vegetation models, or study plant functional trait variation across taxonomic and environmental scales.
- Citations
- 18
- FWCI
- 70.70
- Field percentile
- top 1%
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Understanding the variation in wood densities of trees and its implications for carbon assessments
K. Teegalapalli, C. K. Pandey, … M. Sankaran·bioRxiv·Jan 19, 2019
- A global map of wood density
Fabian Jörg Fischer, Jérôme Chave, … Pablo R. Stevenson·bioRxiv·Aug 25, 2025
- New formula and conversion factor to compute tree species basic wood density from a global wood technology database.
G. Vieilledent, F. J. Fischer, … J. Gerard·bioRxiv·Jul 1, 2018
- Wood structure explained by complex spatial source-sink interactions
A. D. Friend·bioRxiv·May 2, 2020
- Woody species with high wood density are more vulnerable to global change in South China
G. C. Li, S. D. Veresoglou·bioRxiv·Nov 6, 2023
- Just how big is intraspecific trait variation in basidiomycete wood fungal fruit bodies?
S. K. Dawson, M. Jonsson·bioRxiv·Aug 26, 2019