henhaobio
Sign in

NatureEcologyPeer-reviewed journal article

Land-use change undermines the stability of avian functional diversity

Thomas L. WeeksiD, Patrick A. WalkdeniD, David P. EdwardsiD, Alexander Charles LeesiD, Alex L. PigotiD, Andy PurvisiD, Joseph A. TobiasiD

PDFDOIPreprint
In-depth full-text digest available · Jump to digest ↓

Abstract

Land-use change causes widespread shifts in the composition and functional diversity (FD) of species assemblages, yet the impacts on ecosystem resilience remain uncertain. Stability of ecosystem functioning may increase after land-use change because the most sensitive species are removed, leaving more resilient survivors (Balmford, 1996; Clavel et al., 2010; McKinney & Lockwood, 1999). Alternatively, ecosystems may be destabilized if land-use change reduces functional redundancy, accentuating the ecological impacts of further species loss (Fonseca & Ganade, 2001; McCann, 2000). Current evidence is inconclusive, partly because trait data have not been available to quantify functional stability at sufficient scale. Here, we use morphological measurements of 3696 bird species to estimate shifts in functional redundancy following recent anthropogenic land-use change at 1281 sites worldwide, and then assess the sensitivity of these altered assemblages to future species loss. We find that the proportion of disturbance-tolerant species increases after land-use change, but that this does not increase stability. Instead, we show that functional redundancy is reduced and that further species loss will destabilize ecosystem function because relatively few additional extinctions lead to accelerated losses of FD, particularly in trophic groups delivering important ecological services such as seed dispersal and insect predation. Our analyses reveal that land-use change may have major undetected impacts on the stability of key ecological functions, hindering the capacity of natural ecosystems to absorb further declines in functionality caused by ongoing perturbations.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (19 Jan 2025) · CC BY 4.0

Across 1281 bird assemblages and 3696 species, land-use change lowers functional redundancy and resistance to simulated extinctions despite lower vulnerability, hitting frugivores and invertivores hardest.

Prior land-use studies often report only modest functional diversity (FD) losses once species turnover is considered, but a whole-assemblage FD snapshot cannot show the state of individual functions or how stable they are if more species are lost. This work measures functional redundancy, vulnerability and resistance to simulated extinctions in bird assemblages worldwide to test whether modified landscapes are more robust or more fragile.

Key results

  1. ShownFD changes little or slightly under low-intensity human use but declines consistently in more heavily disturbed land-use types, with the steepest loss in intensely urbanized landscapes.

    177 pristine primary vegetation assemblages compared with 1104 in nearby modified landscapes. FD rose slightly in disturbed primary vegetation (P = 0.043) and fell slightly in mature secondary vegetation (P = 0.046). It dropped most in highly urbanized sites (P < 0.001). · Abstract; Statistical analyses

  2. ShownFunctional redundancy (trait space shared by co-occurring species) first rises with light disturbance, then falls significantly under intensive land use, especially in cropland and intense urban areas.

    Redundancy rose in disturbed primary vegetation (P = 0.012). Declines in mature (P = 0.288) and intermediate-age (P = 0.122) secondary vegetation were not significant. Sharp falls in cropland and intense urban (both P < 0.001). Same pattern when intraspecific variation came from 18183 measured individuals (mean 5 per species). · Functional redundancy; Calculation of functional diversity and…

  3. ShownFunctional vulnerability (FV: how strongly the most sensitive species also supply unique function) falls after land-use change, consistent with filtering out sensitive, functionally distinct species.

    Every anthropogenic land-use type except secondary vegetation showed a significant decrease in at least one FV metric. Trait-based and rarity-based FV (rarity model n = 1278 assemblages) declined similarly, relative to pristine primary vegetation. · Functional vulnerability; Functional vulnerability; Figure legends

  4. SuggestedAssemblage-wide redundancy outweighs vulnerability in setting resistance to further species loss. Resistance therefore drops in human-modified landscapes, most in agricultural and urban settings, even though vulnerability is lower.

    Bivariate linear models over 1281 assemblages: redundancy related positively and FV negatively to area-under-curve resistance, for both trait-based and rarity-based removal orders (each P < 0.001). The FV effect was relatively weak. Ties were randomized over 100 iterations, and the half-life (t1/2) metric gave similar results. · Future extinctions and the stability of…; Calculating functional resistance; Statistical analyses; Figure legends

  5. ShownGuild responses diverge. Generalist and granivore FD, redundancy and resistance stay steady or rise in farmland and cities, while frugivores (seed dispersers) and invertivores (insect predators) decline sharply. Whole-assemblage FD and redundancy losses are steeper at tropical latitudes.

    Each guild had its own mixed model: 617 frugivore, 1264 invertivore, 1051 granivore and 1256 generalist assemblages. A latitude split compared 613 tropical with 668 non-tropical sites. Effects are coefficients with 95% confidence intervals against a primary vegetation baseline. · Functional redundancy; Ecological and geographical factors regulating…; Calculation of functional diversity and…; Figure legends

Methods, data and reproducibility

Assemblages (PREDICTS plus Amazonian/Bornean surveys) were placed in trait probability density hypervolumes from diet proportions and four morphological axes (three from a two-step PCA, plus log Hand-wing Index); FD, redundancy, a sensitivity-redundancy vulnerability score and resistance to sensitivity-ordered simulated extinctions were modelled against land use via mixed models.

  • PREDICTS database bird surveys comparing avian diversity across stages of land-use change (starting set before quality filtering) · birds (Aves) · n = 145 studies at baseline; 39 surveys removed
  • Intensive bird surveys from Amazonian and Bornean rainforests, restructured to match the PREDICTS nested study/block/site design · birds (Aves) · n = 17 studies condensed into seven
  • Final global assemblage dataset spanning 10 land-use types (pristine primary vegetation through intense urban) on six continents · birds (Aves) · n = 1281 assemblages from 98 studies
  • Pristine primary vegetation baseline assemblages, mostly forest with some grassland/shrubland · birds (Aves) · n = 177 (152 forest, 25 non-forest)
  • AVONET species-mean morphometrics (seven biometric traits plus Hand-wing Index) · birds (Aves) · n = 3696 species; 41515 individuals measured
  • AVONET individual-level measurements used to estimate intraspecific variation directly in the sensitivity analysis · birds (Aves) · n = 18183 individuals (mean 5 per species)
  • Diet proportions across nine resource types (Pigot et al. 2020, built on EltonTraits) plus AVONET trophic niche classes · birds (Aves)
  • BirdLife International (2021) range polygons used to build genus-level pseudo-species trait averages · birds (Aves) · n = 73 pseudo-species in 133 assemblages
Methods & tools
TPD R package (trait probability densities, TPDMeans), gawdis R package (distances for grouped and proportional traits), ks R package (Hpi.diag plug-in bandwidth estimator), Two-step principal component analysis of morphometric traits, Linear mixed-effects models (study and study-block random effects), Moran's I spatial autocorrelation test, Sensitivity-ordered extinction simulations (area under curve and half-life), PREDICTS database, AVONET trait database, EltonTraits diet dataset, BirdLife International range polygons, Cubic smoothing spline for average extinction curves
Reproducibility gaps
The data and code statements give placeholder repository links (github.com/tomlweeks1994/xxxx) to be completed on acceptance. Details of survey restructuring, land-use classification and sensitivity scores are left to the Supplementary Information.
Limitations (6)
  • The nested study/block structure made it impractical to build spatial autocorrelation into the global models. Instead, autocorrelation was checked separately within studies and blocks using Moran's I, which the authors report did not alter results. · stated by the authors
  • Area-under-curve resistance depends on the order of species loss: removing one morphologically unique species early causes a steep initial FD drop. This motivated the alternative half-life metric. · stated by the authors
  • Hypervolumes need at least four species, so guild coverage is uneven (frugivores in only 617 of 1281 assemblages), and resistance could not be computed for every guild at every site. · stated by the authors
  • FV can overstate future stability because it ignores how redundancy is spread across the whole assemblage. Morphology-diet links are also noisy in some taxa, which is why diet classifications were added. · stated by the authors
  • Extinction scenarios use fixed orderings by trait-based or rarity-based sensitivity scores. Resistance estimates therefore depend on these assumed loss sequences, not on observed future local extinctions. · noted in this summary
  • Inference is space-for-time: modified sites are compared with nearby primary sites rather than following the same assemblages through time. Survey methods and sampled area also differ among studies, and aggregation plus random effects may only partly absorb this. · noted in this summary

Read this if you study how land-use change affects ecosystem function and want a trait-hypervolume framework that links functional redundancy, vulnerability and resistance to simulated extinctions in birds.

Citations
21
FWCI
13.41
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISSoares '22Stewart '20Gomez '20Ceulemans '21Pitzer '26Davrinche '26
74% semantic matchEcology
F. C. Soares, R. F. de Lima, J. M. Palmeirim+2 more·Global Ecology and Biogeography·Mar 30, 2022