Leveraging Remote Sensing and Crowd-Sourced Biodiversity Data for Enhanced Plant Functional Trait Mapping

Description

Technical info (English)

Technical Description of Global Plant Functional Trait Maps (SLA, LNC, LPC)

This dataset provides high-resolution (1 km spatial resolution) global maps of three essential plant functional traits: Specific Leaf Area (SLA), Leaf Nitrogen Concentration (LNC), and Leaf Phosphorus Concentration (LPC). Each trait map includes detailed spatial distributions and statistical metrics derived from remotely sensed data, global biodiversity records (GBIF), and plant trait data (TRY database).

Traits and Measurement Units:

• Specific Leaf Area (SLA): Ratio of leaf area to leaf dry mass ln(mm² mg⁻¹).

• Leaf Nitrogen Concentration (LNC): Ratio of leaf nitrogen mass to leaf dry mass ln(mg g⁻¹).

• Leaf Phosphorus Concentration (LPC): Ratio of leaf phosphorus mass to leaf dry mass ln(mg g⁻¹).

Data Layers Provided:

For each trait (SLA, LNC, LPC), four distributional layers are provided:

1. Community-Weighted Mean (CWM) (band 1): Mean trait value weighted by species abundance within each 1 km pixel in ln-transformed units.
2. Ln-transformed Standard Deviation (SD) (band 2): Within-pixel trait variability on the ln scale; interpreted multiplicatively as exp(±std) around the geometric mean.
3. Skewness (band 3): Asymmetry of the within-pixel trait distribution; positive values indicate right-tailed distributions (e.g., strong environmental filtering), negative values indicate left-tailed distributions.
4. Excess Kurtosis (kurtosis − 3, band 4): Peakedness and tail weight of the within-pixel trait distribution; positive values reflect concentrated, peaked distributions, negative values reflect flat, uniform distributions associated with higher functional evenness.

Quality Data Layers Provided (QA files):

Two spatial diagnostics are provided alongside the trait maps to help users assess local reliability and identify regions where uncertainty may be elevated.

1. Optimal sampling quadrat size layer (band 1): Per-pixel record of the smallest GBIF sampling quadrat (0.04–3 ×10³ km²) achieving ≥80% species coverage completeness, indicating the spatial support required to characterize local plant assemblages.

2. Abundance-weighted TRY–GBIF representativeness layer (band 2): Per-pixel percentage (0–100%) of GBIF-estimated community abundance associated with species lacking trait data in TRY, indicating regions of elevated uncertainty in trait estimates.

Methodology:

Trait maps were generated by integrating three primary data streams:

• GBIF occurrence records (>1 billion geolocated observations): used to estimate species abundances via an adaptive multiresolution sampling approach (0.04–3 ×10³ km² quadrats, ≥80% Chao–Jost completeness threshold), correcting for spatially uneven sampling effort in crowd-sourced biodiversity data.
• TRY Plant Trait Database (>250,000 species): provided species-level mean trait values (SLA, LNC, LPC) in natural-log transformed space.
• Remote sensing-derived fractional PFT composition: downscaled from MODIS (MCD12Q1) using Landsat-informed ML-methods, used to mitigate GBIF sampling biases and as weights to aggregate trait moments into final pixel-level estimates.

Community-weighted statistical moments (mean, standard deviation, skewness, and excess kurtosis) were computed iteratively per PFT and aggregated across PFTs using fractional PFT cover as weights, yielding the final 1 km global trait maps.

Applications:

This dataset enables improved modeling and ecological assessments, including biodiversity predictions, ecosystem functioning, species coexistence analysis, and responses to global environmental changes.

Spatial Coverage:

Global

Temporal Coverage:

Trait data represent current ecological conditions based on the most recent biodiversity and remote sensing datasets.

Resolution:

Spatial: 1 km

Data Format:

Raster (GeoTIFF)