Data and code from: Drivers of tropical rainforest biomass in Papua New Guinea: Large trees and taxonomic diversity dominate over other drivers
Data files
Sep 30, 2026 version files 4.74 MB
-
Code_and_Data.zip
4.72 MB
-
README.md
20.45 KB
Abstract
Tropical forests are among the most diverse and productive ecosystems globally, with aboveground biomass (AGB) storing a substantial portion of terrestrial carbon. While biodiversity, composition, and forest structure have been shown to affect AGB, their relative roles, interactions, and direct and indirect effects, with consideration of tree-size inherent components, remain insufficiently understood, especially in highly diverse old-growth forests. Using data from a 50-ha ForestGEO plot in primary lowland rainforest in Papua New Guinea, we assessed relative importance and individual direct and indirect effects of taxonomic, functional, and phylogenetic diversity, species composition, community trait composition, forest structure, and topography on AGB. Forest structure, represented by the number of large trees, was the dominant predictor of AGB, explaining ~63 % of its variation, but its effect was greatly reduced after correction for tree-size structure. Communities with more conservative trait composition had higher AGB, although this relationship was not retained after size correction. Taxonomic diversity was the next most influential group, while phylogenetic and functional diversity explained little variation in AGB. The taxonomic and phylogenetic diversity affected AGB primarily indirectly through their influence on community trait composition and forest structure. Functional diversity and species composition had negligible effects. Topography shaped AGB indirectly via its influence on diversity and community trait composition. Overall, AGB was highest in subplots that were structurally complex with stands dominated by large trees, high species richness and more conservative community strategies. Our study shows that forest structure, particularly the presence of large trees, is the key determinant of AGB, with biodiversity effects acting mainly through shifts in community trait composition and structure. These findings highlight the importance of conserving not only species richness but also structural complexity and functional trait distributions to maintain carbon storage in tropical forests.
The "folder Code and Data" (Code_and_Data.zip) contains the analysis workflow used to quantify drivers of aboveground biomass (AGB) in the Wanang (ForestGEO) 50-ha tropical forest plot, including (i) a size-corrected biomass null model, (ii) PCA-based data reduction for trait, topography and species composition, (iii) boosted regression trees (BRT) to estimate relative influence of predictor groups, (iv) piecewise structural equation models (SEM) to evaluate direct and indirect effects, (v) Sensitivity analyses to assess the robustness of the results, (vi) additional calculations and graphs used in the main text and supporting information and information about the provided data.
Overview of workflow
- Size-corrected biomass (null model)
- Constructs a null expectation of subplot AGB by randomly re-assigning DBH values within each subplot while keeping species abundances per subplot fixed (sampling without replacement).
- The null AGB mean per subplot is used to obtain size-corrected biomass residuals (AGB_resid) from a regression of observed AGB on null AGB.
- PCA
- Performs separate PCAs for:
- Trait composition (CWM): CWM.DBH, CWM.PVH, CWM.WD, CWM.SLA, CWM.fruit
- Topography: elevation, convexity, slope (meanelev, convex, slope)
- Species composition: community matrix (subplot × species)
- PCA scores are saved and later used in BRT and SEM.
- Performs separate PCAs for:
- Boosted regression trees (BRT)
- Runs BRT models for two responses:
- Observed biomass (AGB.all_species)
- Size-corrected biomass (AGB_resid)
- Forest structure is represented in two alternative ways:
- Large trees: NOLarge_40 (number of trees above a DBH threshold)
- Structural diversity: Gini_DBH
- Each BRT is repeated 100 times, and mean relative influence is reported.
- An additional analysis is performed without grouping variables into logical groups, i.e. with all variables analyzed separately.
- Runs BRT models for two responses:
4. Piecewise SEM
- Fits piecewise SEMs for observed biomass and size-corrected biomass, using PCA axes and diversity predictors.
- Checks multicollinearity using VIF.
5. Sensitivity analyses
A new Sensitivity analyses folder contains three additional analyses used to assess the robustness of the results:
- Semivariograms.R examines residual spatial autocorrelation for all SEM component models using semivariograms with 95 % confidence intervals.
- SEMSpatialCorrected.R refits the SEMs using GLS models with an exponential spatial correlation structure, with spatial range and nugget estimated separately for each component equation.
- DifferentSpatialScale.R compares the main SEMs at the original 20 × 20 m grain and after aggregation into non-overlapping 40 × 40 m blocks, for both observed and size-corrected biomass.
6. Descriptive statistics and graphs
Data and file structure
The Dryad deposit contains three processed data files used in the analyses. All data files distributed in this Dryad deposit are released under the Dryad CC0 waiver and can be reused without contacting the authors. The contact information given below applies only to additional source files that are not included in the Dryad deposit.
The tab-delimited files were written from R with row names. The leftmost unlabeled field is therefore a sequential R row-name index only and has no biological or analytical meaning. Similarly, the first unlabeled field in pca.scores.wan2.csv is a sequential row index.
plot.data.fin.Wan2.txt
Subplot-level data for the 20 × 20 m quadrats. Trait completeness filtering used for the PCA and subsequent analyses is described in PCA.R.
| Variable | Description | Unit / scale |
|---|---|---|
| QuadratName | Unique identifier of the 20 × 20 m subplot | identifier |
| plot | Sequential subplot index in the Wanang plot | identifier |
| SR | Species richness; number of species/mnemonics among living stems in the subplot | count |
| SR_even | Species evenness calculated as Shannon diversity divided by ln(species richness) | unitless, 0–1 |
| NOI | Number of living tree individuals/stems in the subplot | count |
| Simpson | Simpson concentration/dominance index, calculated as sum of squared relative abundances | unitless; lower values indicate greater diversity |
| NOLarge_60 | Number of living stems with DBH > 60 cm | count |
| NOLarge_40 | Number of living stems with DBH > 40 cm | count |
| NOLarge_20 | Number of living stems with DBH > 20 cm | count |
| NOLarge_10 | Number of living stems with DBH > 10 cm | count |
| NA_prop.DBH | Proportion of individuals lacking the species-level DBH trait value used for CWM calculation | proportion, 0–1 |
| CWM.DBH | Abundance-weighted community mean of species-level mean diameter at breast height (DBH) | cm |
| NA_prop.PVH | Proportion of individuals lacking plant vegetative height data | proportion, 0–1 |
| CWM.PVH | Abundance-weighted community mean of plant vegetative height (PVH) | m |
| SD_PVH | Standard deviation of log10-transformed species-level PVH values for species present in the subplot | log10 scale |
| Gini_PVH | Gini coefficient of species-level PVH values for species present in the subplot | unitless, 0–1 |
| NA_prop.WD | Proportion of individuals lacking wood-density data | proportion, 0–1 |
| CWM.WD | Abundance-weighted community mean wood density (WD) | g cm -3 |
| NA_prop.SLA | Proportion of individuals lacking specific leaf area data | proportion, 0–1 |
| CWM.SLA | Abundance-weighted community mean specific leaf area (SLA) | mm2 mg -1 |
| NA_prop.LDMC | Proportion of individuals lacking leaf dry matter content data | proportion, 0–1 |
| CWM.LDMC | Abundance-weighted community mean leaf dry matter content (LDMC) | mg g -1 |
| NA_prop.fruit | Proportion of individuals lacking fruit-weight data | proportion, 0–1 |
| CWM.fruit | Abundance-weighted community mean fruit weight | g |
| NA_prop.CN | Proportion of individuals lacking leaf carbon:nitrogen ratio data | proportion, 0–1 |
| CWM.CN | Abundance-weighted community mean leaf carbon:nitrogen ratio | unitless ratio |
| NA_prop.rao | Proportion of individuals lacking one or more traits used to calculate multivariate functional diversity (SLA, PVH, WD and species-level mean DBH) | proportion, 0–1 |
| RaoQ | Rao's quadratic entropy calculated from SLA, PVH, WD and species-level mean DBH using FD::dbFD |
unitless functional-diversity index |
| FEve | Functional evenness calculated using FD::dbFD |
unitless, 0–1 |
| SR.filtered | Species richness after taxonomic matching/filtering for the phylogenetic analysis | count |
| PD | Faith's phylogenetic diversity; sum of branch lengths after square-root transformation of phylogenetic branch lengths | transformed phylogenetic branch-length units |
| MPD | Unweighted mean pairwise phylogenetic distance among species in the subplot, using square-root-transformed phylogenetic distances | transformed phylogenetic distance units |
| PSE | Phylogenetic species evenness returned by picante::pse |
unitless |
| NA_prop.AGB | Proportion of individuals excluded from the individual-tree AGB calculation because required trait information was unavailable | proportion, 0–1 |
| AGB | Summed aboveground biomass of stems for which individual-tree AGB could be calculated | kg dry mass per 20 × 20 m subplot |
| DBHshift_na_rem | Absolute difference in mean DBH before and after excluding stems lacking required trait data for AGB calculation | cm |
| AGB.all_species | Subplot AGB extrapolated from calculated stems to the full living-tree community | kg dry mass per 20 × 20 m subplot |
| BA | Total basal area of living stems in the subplot | cm2 per 20 × 20 m subplot |
| SD_BA | Standard deviation of log10-transformed individual basal area | log10 scale |
| SD_DBH | Standard deviation of log10-transformed individual DBH | log10 scale |
| mean_DBH | Mean DBH of living stems | cm |
| Gini_BA | Gini coefficient of individual basal area | unitless, 0–1 |
| Gini_DBH | Gini coefficient of individual DBH; used as an alternative measure of forest structural diversity | unitless, 0–1 |
| AGB_null_mean | Mean expected subplot AGB across null-model simulations | kg dry mass per 20 × 20 m subplot |
| AGB_null_sd | Standard deviation of expected subplot AGB across null-model simulations | kg dry mass per 20 × 20 m subplot |
| AGB_SES | Standardized effect size derived from the AGB null model | unitless |
| AGB_resid | Size-corrected biomass: residual from the regression of log10 observed AGB (AGB.all_species) on log10 expected AGB (AGB_null_mean) | residual on log10 scale |
| meanelev | Mean elevation of the subplot | m a.s.l. |
| convex | Topographic convexity; positive values indicate locally convex terrain and negative values locally concave terrain | m |
| slope | Terrain slope | degrees |
| startx | Local x-coordinate of subplot centre within the 50-ha plot | m |
| starty | Local y-coordinate of subplot centre within the 50-ha plot | m |
pca.scores.wan2.csv
PCA site scores used in the BRT and SEM analyses.
| Variable | Description | Unit / scale |
|---|---|---|
| first unlabeled column | Sequential R row index; not an analytical variable | index |
| CWM.pca1 | Site score on PC1 of the PCA of community-weighted mean DBH, PVH, WD, SLA and fruit weight | unitless PCA score |
| CWM.pca2 | Site score on PC2 of the same community-trait PCA | unitless PCA score |
| top.pca1 | Site score on PC1 of the PCA of mean elevation, convexity and slope | unitless PCA score |
| top.pca2 | Site score on PC2 of the same topographic PCA | unitless PCA score |
| comp.pca1 | Site score on PC1 of the PCA of subplot species composition | unitless PCA score |
| comp.pca2 | Site score on PC2 of the PCA of subplot species composition | unitless PCA score |
| QuadratName | Unique identifier of the corresponding 20 × 20 m subplot | identifier |
indiv.level.biomass.txt
Derived individual-tree data used mainly for descriptive analyses and supplementary figures.
| Variable | Description | Unit / scale |
|---|---|---|
| first unlabeled field | Sequential R row-name index; not an analytical variable | index |
| QuadratName | Unique identifier of the 20 × 20 m subplot | identifier |
| Plot | Sequential subplot index in the Wanang plot | identifier |
| Mnemonic | Species mnemonic/code used in the Wanang census | categorical species code |
| DBH | Taper-corrected diameter at breast height (equivalent diameter at 1.3 m) | cm |
| WD_merged_fin | Species-level wood density used for individual-tree biomass estimation | g cm -3 |
| PVH_merged_fin | Species-level plant vegetative height associated with the tree species | m |
| AGB | Estimated aboveground biomass of the individual tree using Chave et al. (2014), equation 7 | kg dry mass |
Additional source files not included in the Dryad deposit
Wanang ForestGEO tree census (Wanang2.fin.txt)
The original Wanang tree census file, Wanang2.fin.txt, is not included in this Dryad deposit. Wanang2.fin.txt is the original individual-tree census file from which several of the processed data products provided in this Dryad deposit were derived. The processed data required to reproduce the analyses presented in the manuscript are included in the Dryad archive; the original census file is only needed to regenerate these derived datasets from the raw tree-level data.
Wanang plot data can be requested through the ForestGEO data request portal: http://ctfs.si.edu/datarequest/
For use of this external tree census, we ask users to make a courtesy request to the Wanang ForestGEO principal investigators Vojtech Novotny (novotny@entu.cas.cz) and George Weiblen (gweiblen@umn.edu).
This courtesy request applies only to the external Wanang2.fin.txt file and does not apply to any files distributed in this Dryad deposit. All files deposited in Dryad are available under the Dryad CC0 waiver.
Full trait compilation (Traits.txt)
The full species-level trait compilation used by Null_model.R, Traits.txt, is not included in this Dryad deposit. The compiled trait file contains data assembled from multiple sources and can be obtained by contacting Aleš Lisner (lisnea00@jcu.cz).