Traceable Claims in Ecology

 Quick Analysis Plan Copied

 Long Analysis Plan

Evidence base first: the eight supplied records contain the methodological machinery for this analysis — multivariate ordination, NDVI/NMRI/fractional-cover remote sensing, variance partitioning/SEM, and dynamic vegetation modelling — but none reports dark-diversity data itself, and the primary dark-diversity literature is absent. Species-pool specifics below are therefore a blueprint to be sourced, not cited values.

1 · Core computation: species pools and dark diversity

Assemble a sites × species presence matrix from natural-vegetation plots, build regional co-occurrence pools, score suitability of absent species, and define dark diversity as suitable-but-absent species per site. Minimal working core:

import numpy as np
# X: sites x species presence (0/1) matrix
C = X.T @ X; np.fill_diagonal(C, 0)         # co-occurrence counts
P = C / np.maximum(C.sum(0), 1)             # conditional association
S = X @ P                                   # site suitability (Beals-type)
pool = (X == 0) & (S > np.median(S, axis=1, keepdims=True))
dark = pool.sum(1)                          # dark diversity per site
ratio = dark / np.maximum(X.sum(1), 1)      # impoverishment index

Known unknown: the suitability threshold and pool definitions cannot be cited from the supplied records and must be sourced before running.

2 · Covariate layers with provenance

• Greenness: NDVI time series; across 50 Yangtze Delta cities, degradation hotspots tracked economic and demographic drivers more than urban land expansion NDVI urbanization regression, 50 Yangtze cities.
• Water status: site-normalized GNSS-reflectometry NMRI; cross-site NMRI–NDVI correlation exceeded 0.7 across four Montana grassland sites over 46 visits GNSS-reflectometry NMRI validation, Montana grasslands.
• Cover fractions: NDVI plus cellulose-absorption-index unmixing (Hyperion upscaled to MODIS) separating photosynthetic, non-photosynthetic, and bare-soil fractions Fractional cover unmixing in Australian savannas.
• Disturbance: Sentinel-2 multitemporal indices with random forest; Leaf Chlorophyll Content Index and NDRE-SWIR2 best tracked post-fire recovery Post-fire recovery indices, Ecuadorian Andes.

3 · Attribution and ordination toolkit

Reuse the Tibetan Plateau framework (Mann-Kendall trends, variance partitioning, LMG importance, SEM) to attribute dark diversity to climate, edaphic, and land-use covariates. Its headline result rejects temperature-only models: non-temperature factors explained about 45% of active-layer thickness and top-of-permafrost variation versus roughly 18–22% for air temperature, and SEM gave vegetation cover a −38.8% direct effect on permafrost state Tibetan permafrost driver attribution framework. Ordination is equally necessary: at two Vancouver Island marshes, elevation explained under half of vegetation variation, and dyking rewrote edaphic filters — organic content diverged by 27.000000000000004 points at Kokish (n=26/26) yet only -1.6999999999999993 points at Cluxewe (n=27/25) Dyked vs undyked marsh ordination, Vancouver Island. GAMs on 1,920 one-metre arctic-alpine plots add nonlinear soil-temperature and moisture response surfaces GAM analysis of 1,920 arctic-alpine plots.

Values as reported (approximate); categories may overlap within the 45% non-temperature total.

Reported means; Kokish shows extreme dyking divergence, Cluxewe does not.

4 · Projection, validation, falsification

Project future pool shifts with individual-based dynamic vegetation modelling integrating fire, phenology, and carbon allocation, as in aDGVM across African grassland–savanna–forest mosaics aDGVM simulation of African vegetation. Validate with held-out plots, per-biome cross-validation, and sensitivity analysis across pool thresholds. Falsification test: if estimated dark diversity is randomly distributed across land-use and socio-economic gradients — unlike Yangtze degradation, which tracked them — the anthropogenic-impoverishment hypothesis fails NDVI urbanization regression, 50 Yangtze cities.

5 · Blind spots

• No dark-diversity primary literature or global plot dataset exists in the supplied records; pool estimation is unverified here.
• Point plots versus gridded climate products create scale mismatch, flagged by the Tibetan authors themselves Tibetan permafrost driver attribution framework.
• NMRI validation covered one growing season at four drought-affected sites GNSS-reflectometry NMRI validation, Montana grasslands; aDGVM omits soil-carbon dynamics and vegetation–climate feedbacks aDGVM simulation of African vegetation; the 1986 marsh analysis used linear techniques only Dyked vs undyked marsh ordination, Vancouver Island.

Run AI Scientist AnalysisBest Evidence: Dark Diversity MethodsMeta-analysis: Dark Diversity Drivers Evolve Bioinformatics/Coding/Analysis Plan