Dark diversity in temperate forests of northeastern China: drivers and implications
Data files
Aug 27, 2025 version files 106.40 KB
-
code.R
4.33 KB
-
data_VPA_RF_SAR.csv
97.64 KB
-
README.md
4.43 KB
Abstract
Understanding the species composition of a community, including those species present and those absent but potentially able to occur, is vital for assessing biodiversity changes and informing conservation planning. Typically, studies focus on observed taxonomic diversity but ignore undetected species expected to be present based on co-occurrence patterns—referred to as dark diversity. Dark diversity serves as a sensitive indicator of biodiversity change, often responding earlier than observed diversity. However, its underlying drivers, especially environmental and anthropogenic factors, remain poorly understood in forest ecosystems. In this study, we quantified both dark and observed diversity and applied logistic regression to identify traits influencing species’ likelihood of belonging to dark diversity at the species level. Variance partitioning and spatial autoregressive models were used to disentangle the effects of environmental and human drivers at the plot level. Our results revealed that key traits such as mycorrhizal type, specific leaf area and tree height determine a species’ likelihood of belonging to dark diversity, and that dark and observed diversity respond differently to environmental and anthropogenic factors. Dark diversity was mainly influenced by the interaction of environmental factors and human impacts, with annual mean temperature being the strongest environmental driver. Observed diversity, in contrast, was most influenced by annual mean precipitation. Dark diversity increased with mean annual temperature, declined under the combined influence of human footprint and precipitation, and increased with the interaction between human footprint and precipitation seasonality. Observed diversity was positively affected by precipitation and altitude, and negatively influenced by human footprint-temperature and altitude-roadless area interactions. Our findings highlight the importance of addressing climate and human pressures in conservation planning, as dark diversity signals hidden constraints on species establishment and persistence.
https://doi.org/10.5061/dryad.3n5tb2rvc
Description of the data and file structure
This dataset contains information from 456 0.1-hectare forest plots sampled in northeastern China. It includes measures of biodiversity, climate, soil, topography, and human influence for each plot. This repository contains the code (code.R) for the analysis focusing exclusively on the plot-level drivers (i.e., what factors influence DD and OD values per plot). It employs a multi-method statistical approach to partition the variance explained by different groups of drivers and to model their effects while accounting for spatial autocorrelation.
Core Analyses:
- Variance Partitioning: Separates the influences of environmental variables, spatial structure (PCNM vectors), and human factors on DD and OD.
- Partial RDAs: Tests the unique statistical significance of each variable group.
- Random Forest Models: Ranks the importance of individual environmental predictors.
- Spatial (SAR) Models: Models key interaction effects between climate and human factors while accounting for spatial autocorrelation.
Requirements:
- Data: A dataframe named data_VPA_RF_SAR must be pre-loaded.
- Libraries: Requires the vegan, randomForest, and spdep packages.
- Spatial Settings: The SAR model uses a spatial weights matrix based on the 4 nearest neighbors (k=4).
Files and variables
File: data_VPA_RF_SAR.csv
Description:
Variables
- plot: A unique identifier for each of the 456 sampled forest plots in the study area. Categorical (Plot ID)
- DD: Dark Diversity: The pool of species absent from a local site but present in the surrounding region and potentially able to inhabit the local site conditions. Number of species (count)
- OD: Observed Diversity: The number of species actually recorded and observed within the forest plot. Number of species (count)
- longitude: The geographic longitude coordinate of the plot's center. Decimal degrees (WGS 84)
- latitude: The geographic latitude coordinate of the plot's center. Decimal degrees (WGS 84)
- BIO1: Annual Mean Temperature: The average annual temperature at the plot location. Degrees Celsius (°C)
- BIO12: Annual Mean Precipitation: The total annual precipitation at the plot location. Millimeters (mm)
- BIO15: Precipitation Seasonality: The coefficient of variation (CV) of monthly precipitation totals. Measures how precipitation fluctuates throughout the year. Unitless (Coefficient of Variation)
- altitude: The elevation of the ground surface at the plot's center above sea level. Meters (m)
- slope: The incline of the terrain at the plot, likely representing the angle of the surface from horizontal. Degrees (°)
- moisture: Soil Volumetric Water Content: A measure of the amount of water present in the soil sample from the plot. (-33 to -1500 kPa, cm m-1)
- C_N: Carbon to Nitrogen Ratio: A key soil property indicating the ratio of organic carbon mass to nitrogen mass in the soil. Ratio (mass carbon / mass nitrogen)
- pH: Soil pH: A measure of the acidity or alkalinity of the soil sample. pH scale (0 - 14)
- HF: Human Footprint: A cumulative measure of direct human pressure on the environment (e.g., from infrastructure, land use, population density). Index (typically 0-50 or 0-100, where 0 = no pressure)
- HC: Human Footprint Change: The change in the Human Footprint index over a specific time period. Likely indicates intensification or reduction of human pressure. Change in Index value over time (e.g., ΔHF/decade)
- RA: Roadless Area: An indicator of the plot's proximity to or location within areas devoid of roads. (km2)
- PA: Protected Area: An indicator of whether the plot is located within a legally designated protected area. (km2)
- PCNM_X7: PCNM Vector 7: A spatial eigenvector from a Principal Coordinates of Neighbourhood Matrix (PCNM) analysis, representing a specific scale of spatial patterning. Unitless (Eigenvector score)
- PCNM_X14: PCNM Vector 14: A spatial eigenvector from a PCNM analysis, representing a specific scale of spatial patterning. Unitless (Eigenvector score)
- PCNM_X115: PCNM Vector 115: A spatial eigenvector from a PCNM analysis, representing a specific (often very fine) scale of spatial patterning. Unitless (Eigenvector score)
