Data and code from: Scale dependence of woody plant β-diversity in a tropical rainforest metacommunity

Published Oct 09, 2026 on Dryad. https://doi.org/10.5061/dryad.1c59zw4bz

Data files

Oct 09, 2026 version files 14.94 MB

ALL_ENV_dbMEM_DATA_mean.csv

8.04 MB
ALL_ENV_dbMEM_DATA_range.csv

5.81 MB
Beta_Div_components_multiN_5to50.csv

1.04 MB
beta_env_analysis_v2.R

36.50 KB
README.md

22.74 KB

Abstract

β-diversity is a central concept to understanding community assembly, yet how its decomposition components and underlying drivers vary with spatial extent remains insufficiently understood. We used a nested metacommunity framework to examine the scale dependence of total β-diversity and its species replacement and richness difference components, and to identify how environmental and spatial predictors contribute across scales. We analyzed woody plant diversity in 250 plots arranged on a contiguous 1×1km grid in a tropical rainforest on Hainan, China. By aggregating nearest-neighbor plots into metacommunities, we constructed a continuous gradient of spatial extent and quantified total β-diversity (BDtotal), species replacement (Repl), and richness difference (RichDif) for each metacommunity realization. β-diversity increased modestly with spatial extent, but its internal structure was remarkably stable: species replacement consistently dominated total β-diversity, whereas richness differences contributed less and changed little across scales. Environmental heterogeneity was positively associated with β-diversity, primarily through replacement rather than richness differences. The variation explained by both environmental predictors and spatial structure increased with extent before approaching a plateau. Spatial structure explained the largest share of variation in BDtotal, Repl, and RichDif, while the relative importance of environmental predictors shifted with scale: soil variables were more influential at finer extents, whereas bioclimatic variables gained importance at broader extents. Environmental predictors were more strongly associated with species replacement than with richness differences, whereas spatial structure explained both components to a similar degree. Overall, increasing spatial extent amplified the magnitude of β-diversity more than it altered the balance between its components, revealing species replacement as a scale-robust feature of tropical forest metacommunities. These findings show that tropical forest β-diversity emerges from the joint effects of environmental heterogeneity and spatial structure, and they highlight the need to conserve spatially dispersed, compositionally complementary habitat mosaics rather than focusing solely on local richness hotspots.

Dataset DOI: 10.5061/dryad.1c59zw4bz

Scale-dependent woody plant beta diversity and environmental/spatial predictors in Bawangling tropical rainforest, China

Dataset overview

This dataset contains derived woody plant beta-diversity metrics and environmental and spatial predictors for metacommunities of varying spatial extent in the Bawangling Nature Reserve, within Hainan Tropical Rainforest National Park, Hainan Island, China. The dataset was developed to investigate (1) changes in total beta diversity and its species-replacement and richness-difference components with spatial extent; (2) relationships between community composition and environmental heterogeneity; and (3) the relative explanatory contributions of climate, soils, and spatial structure across extents.

Woody vegetation was surveyed in 2020 in 250 field plots, each 20 m × 20 m, located in a contiguous network of 1 km × 1 km grid cells (one vegetation plot per grid cell). All woody stems with diameter at breast height (DBH) ≥ 1 cm were identified to species. The present three files contain derived metacommunity summaries, not the original stem-by-stem inventory, plot coordinates, species-by-plot community matrices, or environmental raster layers.

Construction of metacommunities and the spatial-extent index

Each of the 250 vegetation plots was used in turn as a focal plot. For a specified size N, the focal plot and its N − 1 geographically nearest neighboring plots were grouped as one metacommunity realization. Values of N range from 5 through 50, inclusive, giving 46 nested metacommunity sizes for each focal plot and 250 × 46 = 11,500 focal-plot-by-size realizations. A metacommunity of N = 5, for example, comprises the focal plot plus its four nearest neighbors.

N is the number of local plots included in the metacommunity, not a distance in meters or a fixed geographic area. The physical radius varies among focal plots and is defined by the distance to the (N − 1)th nearest neighbor. Metacommunities for nearby focal plots can overlap, and realizations across successive N are nested; consequently, rows and measurements must not be assumed to be statistically independent.

Files and data structure

File	Rows (excluding header)	Columns	Data organization and purpose
`Beta_Div_components_multiN_5to50.csv`	250	231	One row per numerically identified focal metacommunity; beta-diversity metrics and component proportions stored in wide format for every N = 5–50.
`ALL_ENV_dbMEM_DATA_mean.csv`	11,500	48	One row per focal metacommunity and N; arithmetic mean of each environmental variable and spatial eigenvector score across the N member plots.
`ALL_ENV_dbMEM_DATA_range.csv`	11,500	48	One row per focal metacommunity and N; within-metacommunity range (maximum minus minimum) of each environmental variable and spatial eigenvector score across the N member plots.

All files are comma-separated text tables with a header row; decimals use a period (.). No blank/NA cells were detected in the three submitted files. Climate and soil variables are derived from gridded environmental datasets, whereas MEM variables are mathematical spatial predictors derived from plot coordinates. The predictor tables contain 19 bioclimatic, 9 soil, and 18 spatial eigenvector fields in addition to the two identifiers.

File 1: `Beta_Div_components_multiN_5to50.csv`

Row identifier

metaplot: integer identifier from 1 through 250, representing one focal metacommunity series; an identifier, not a response value or a spatial distance.

Repeated column pattern for N = 5, 6, ..., 50

Each N contributes the following five columns, giving 46 × 5 = 230 metric columns in addition to metaplot. Here # stands for the number of local plots, such that BDtotal_N5 is the total beta diversity for the metacommunity formed from the focal plot and its four closest neighboring plots.

Column name pattern	Meaning	Units / interpretation
`BDtotal_N#`	Total beta diversity: Jaccard-based compositional variation among the N local woody plant assemblages, using presence–absence information.	Dimensionless beta-diversity measure; not a percentage.
`Repl_N#`	Species-replacement component of total beta diversity: compositional differences attributed to exchanges of species identities among plots.	Dimensionless, on the same additive scale as BDtotal.
`RichDif_N#`	Richness-difference component of total beta diversity: compositional differences associated with unequal numbers of species among plots.	Dimensionless, on the same additive scale as BDtotal.
`Repl/BDtotal_N#`	Fraction of total beta diversity accounted for by species replacement, calculated as `Repl_N# / BDtotal_N#`.	Dimensionless proportion from 0 to 1; multiply by 100 for percent.
`RichDif/BDtotal_N#`	Fraction of total beta diversity accounted for by richness differences, calculated as `RichDif_N# / BDtotal_N#`.	Dimensionless proportion from 0 to 1; multiply by 100 for percent.

The partition follows a replacement–richness difference framework, not an interchangeable “replacement–nestedness” decomposition. For a given focal plot and N, BDtotal_N# = Repl_N# + RichDif_N#, and the two component proportions sum to 1 (allowing for floating-point rounding). In the submitted data, these equalities hold to numerical precision.

Example: The field Repl/BDtotal_N20 represents the fraction of Jaccard-based total beta diversity arising from species replacement when the focal plot is combined with its 19 nearest neighbors. A value of 0.75 means 75%, not 0.75%.

Files 2 and 3: environmental and spatial predictor tables

Both ALL_ENV_dbMEM_DATA_mean.csv and ALL_ENV_dbMEM_DATA_range.csv have the same column names and the same 11,500 combinations of focal metacommunity and N.

Common identifiers

Column	Meaning	Type / units
`metaplot`	Original alphanumeric code for the focal vegetation plot (for example, `Z10`); it identifies a focal metacommunity series.	Character identifier. Leading letters and numbers are plot codes, not predictor values.
`N`	Number of member plots in the focal metacommunity, including the focal plot itself; integer values 5–50.	Count of plots (unit: plots).

What `mean` and `range` mean

For each numeric predictor X and each (metaplot, N) combination, the tables summarize the values assigned to the N constituent local plots:

Mean file: mean(X) = [sum of X across the N member plots] / N.
Range file: range(X) = max(X across the N member plots) − min(X across the N member plots).

Thus, the mean describes average environmental conditions or average MEM scores of a metacommunity, whereas the range describes within-metacommunity variation. A range of zero means all included plots have equal values for that predictor. Ranges retain the same stored numerical units and scale factors as the corresponding means. Environmental summaries are not direct measurements in each 20 m × 20 m plot; the climate and soil layers are spatial predictions. Model standardization, where used in downstream analyses, is separate from the raw-scale values archived here.

Bioclimatic variables (`bio1`–`bio19`)

The 19 WorldClim bioclimatic variables describe temperature, precipitation, and their seasonality. A quarter means three consecutive months. The temperature values in these CSV files are expressed in degrees Celsius, precipitation in millimeters, and certain seasonality indices retain the WorldClim scale described below. Definitions follow WorldClim: https://www.worldclim.org/data/bioclim.html

Variable	Definition	Units / numerical scale in these CSVs
`bio1`	Annual mean temperature	°C
`bio2`	Mean diurnal temperature range	°C
`bio3`	Isothermality: 100 × (`bio2` / `bio7`)	%
`bio4`	Temperature seasonality: standard deviation of monthly mean temperatures multiplied by 100	100 × °C; divide by 100 to obtain temperature SD in °C
`bio5`	Maximum temperature of warmest month	°C
`bio6`	Minimum temperature of coldest month	°C
`bio7`	Annual temperature range (`bio5` − `bio6`)	°C
`bio8`	Mean temperature of wettest quarter	°C
`bio9`	Mean temperature of driest quarter	°C
`bio10`	Mean temperature of warmest quarter	°C
`bio11`	Mean temperature of coldest quarter	°C
`bio12`	Annual precipitation	mm
`bio13`	Precipitation of wettest month	mm
`bio14`	Precipitation of driest month	mm
`bio15`	Precipitation seasonality (coefficient of variation)	%, in the WorldClim convention
`bio16`	Precipitation of wettest quarter	mm
`bio17`	Precipitation of driest quarter	mm
`bio18`	Precipitation of warmest quarter	mm
`bio19`	Precipitation of coldest quarter	mm

Climate source: WorldClim bioclimatic raster layers, reported spatial resolution approximately 30 arc seconds. Values in these tables represent metacommunity-level summaries of the plot-associated raster values. They are not measurements taken during the 2020 vegetation survey. The specific WorldClim version and climate reference period are not encoded in the three CSV files.

Soil variables (`BD`, `ST`, `CF`, `PH`, `TK`, `TN`, `TP`, `SOC`, `CEC`)

These nine variables were obtained from the National Soil Information Grids of China (approximately 90 m horizontal resolution; Liu et al., 2022). Important: the numerical entries in the archived CSVs retain the source rasters' integer-style scale factors. To recover values in the physical units shown below, divide the stored CSV value by the indicated factor. The same divisor applies to mean and range values.

Variable	Definition	Physical unit after conversion	Divide CSV values by
`BD`	Soil bulk density	g/cm³	1000
`ST`	Soil thickness (soil profile depth)	cm	1
`CF`	Coarse fragments greater than 2 mm in diameter	% by volume	1
`PH`	Soil pH measured in water (pH-H₂O)	pH units (dimensionless)	100
`TK`	Total soil potassium content	g/kg	100
`TN`	Total soil nitrogen content	g/kg	100
`TP`	Total soil phosphorus content	g/kg	100
`SOC`	Soil organic carbon content (concentration, not carbon stock)	g/kg	100
`CEC`	Cation exchange capacity	cmol(+)/kg	100

Conversion examples: A stored PH value of 500 corresponds to a pH of 5.00. A stored BD value of 1100 corresponds to 1.100 g/cm³. A stored SOC value of 2500 corresponds to 25.00 g/kg. A PH range of 40 corresponds to a difference of 0.40 pH units.

Soil source and scale factors: Soil SubCenter, National Earth System Science Data Center, National Science & Technology Infrastructure of China (https://soil.geodata.cn/). The data source's published unit and scale-factor documentation is available at https://www.geodata.cn/data/datadetails.html?dataguid=36810085119113&docId=11537 ; see also https://soil.geodata.cn/ztsj.html and Liu et al. (2022), DOI: https://doi.org/10.1016/j.scib.2021.10.013 . The national soil grids include multiple depth intervals. The specific soil depth interval(s) used for these derived columns are not encoded in the supplied CSV files and should not be inferred from column names.

Spatial-structure variables (distance-based Moran's eigenvector maps: `MEM#`)

The following 18 columns appear in both predictor files:

MEM1, MEM2, MEM3, MEM4, MEM5, MEM6, MEM7, MEM8, MEM10, MEM12, MEM13, MEM14, MEM18, MEM21, MEM23, MEM24, MEM30, MEM32.

These are distance-based Moran's eigenvector map (dbMEM) scores computed from spatial relationships among the vegetation plots. Each axis represents a distinct pattern of spatial autocorrelation; positive- and negative-valued eigenvector scores are possible. The number after MEM identifies the eigenvector axis. Skipped MEM numbers are not missing data cells: those axes are not present in the archived predictor subset. The archived scores are dimensionless.

In the mean file, each MEM# column is the average score of that eigenvector across the N member plots. In the range file, it is the maximum minus minimum score across those plots. Spatial eigenvectors can capture patterns consistent with dispersal limitation, spatial history, or unmeasured spatially structured environments, but they do not by themselves identify a causal mechanism. The MEM range is a range of eigenvector scores, not a geographic radius or distance.

Linking files and identifiers: important limitation

The two environmental/spatial predictor files can be joined directly using the two-column key (metaplot, N). Each file contains one record per combination of 250 alphanumeric focal codes and 46 values of N.

However, the beta-diversity file uses numerical metaplot identifiers (1–250), whereas the predictor files use alphanumeric focal codes (for example, Z10). No identifier correspondence is provided in these three files. Therefore, it is not valid to assume that metaplot = 1 in the beta-diversity file is the same focal plot as the first metaplot row in either predictor file, or to merge the beta-diversity file with predictor data by matching these IDs directly. A separate, verified numeric-to-alphanumeric focal plot crosswalk is required for observation-level joining. The row order alone should not be used to create this mapping without checking the original data-processing scripts or plot metadata.

A future crosswalk should ideally contain columns named beta_metaplot_id (integer 1–250) and env_metaplot_id (original plot code, such as Z10), with one verified mapping per focal plot. No crosswalk has been inferred or fabricated in this documentation.

Scope, interpretation, and practical considerations

Observation unit: a focal-plot-centered metacommunity at a particular N; neither a single stem nor a single 20 m × 20 m plot.
Temporal coverage: the woody plant inventory was conducted in 2020; the gridded climate and soil predictors come from pre-existing spatial products and need not represent field measurements in that year.
Spatial interpretation: N measures the number of constituent plots. Geographic radii and focal coordinates are not included in these three tables.
Measurement and scale: source gridded predictors have coarser spatial support than individual vegetation plots; a numerical summary of raster data should not be interpreted as a 20 m ground measurement or an increase in native raster accuracy.
Statistics: focal realizations are spatially overlapping, and the same focal plot contributes observations at 46 nested extents. Account for this non-independence and spatial autocorrelation in inferential analyses.
Missingness and checks: the supplied three CSVs contain no blank or NA values. All 250 × 46 recorded beta-diversity partition identities are consistent to floating-point precision. The two environmental predictor tables have matching (metaplot, N) keys.
Availability of raw inputs: original woody plant observations, complete geospatial coordinates, raw environmental rasters, and the source soil-depth selections are not included among these three files. Additional source data or processing documentation may be needed to reproduce the original ecological observations or link numeric and alphanumeric focal identifiers.

Brief summary of the archived biodiversity patterns

Averaging the Repl/BDtotal_N# values across all 250 focal series and all 46 extents gives a species-replacement fraction of approximately 0.744 (74.4%), with little change across N = 5–50. Mean BDtotal_N# across the 250 focal series increases from approximately 0.393 at N = 5 to 0.423 at N = 50. These summaries are descriptive of the archived data; metacommunity realizations are overlapping and should not be considered independent experiments. The dataset enables further examination of extent-dependent beta diversity and associations with bioclimatic, edaphic, and spatial predictors, subject to the identifier-linkage limitation above.

Brief summary of the R scripts

The R file "beta_env_analysis_v2.R" was used to perform the statistical analyses underlying the main results and to generate the figures presented in the associated manuscript.

Key source documentation

WorldClim. Bioclimatic variables: https://www.worldclim.org/data/bioclim.html
Soil SubCenter, National Earth System Science Data Center. National Soil Information Grids of China, documentation for raster unit conversion factors: https://www.geodata.cn/data/datadetails.html?dataguid=36810085119113&docId=11537
Liu, F., Wu, H., Zhao, Y., Li, D., Yang, J., Song, X., et al. (2022). Mapping high resolution National Soil Information Grids of China. Science Bulletin, 67, 328–340. https://doi.org/10.1016/j.scib.2021.10.013
Legendre, P., & De Cáceres, M. (2013). Beta diversity as the variance of community data: dissimilarity coefficients and partitioning. Ecology Letters, 16, 951–963.
Legendre, P. (2014). Interpreting the replacement and richness difference components of beta diversity. Global Ecology and Biogeography, 23, 1324–1334.

This README documents the three CSV files listed above. All variable names and filename capitalization correspond to the submitted files.