henhaobio
Sign in

Proceedings of the National Academy of SciencesEcologyPeer-reviewed journal article

Seed dispersal disruption limits tropical forest regrowth

Evan C. FrickeiD, Susan C. Cook‐PattoniD, Charles F. HarveyiD, César TerreriD

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

Abstract

Many trees depend on animals for seed dispersal, and human activities that disrupt seed dispersal by animals may impact forest regeneration and carbon storage. Yet whether expected negative impacts are observable across regrowing forests remains untested. We modeled seed dispersal disruption and its relationship to aboveground carbon accumulation observed across 3026 sites in the tropics, where most trees are animal dispersed. We found that seed dispersal disruption explains wide variation in local carbon accumulation rates. Across areas identified for restoration, we estimate that seed dispersal disruption reduces carbon accumulation rates by 57% on average. These results advance understanding of animal biodiversitys impact on forest carbon and emphasize the need to address biodiversity loss and climate change together.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (12 Dec 2024) · CC BY 4.0

In field records from 3026 tropical sites, modeled bird and mammal seed dispersal disruption predicts slower natural-regrowth carbon accumulation, an estimated 57% average loss at candidate restoration sites.

Local field studies and process-based models (one predicting 38% lower carbon storage in more fragmented Atlantic Forest regrowth) suggested that losing animal seed dispersers should reduce forest carbon, but whether this signal appears across field records of tropical regrowth had not been tested. The authors build a pantropical map of seed dispersal disruption for birds and mammals and link it to observed carbon accumulation, which bears on natural regrowth as a climate mitigation strategy.

Key results

  1. ShownSites with more severe seed dispersal disruption accumulated aboveground carbon more slowly during natural regrowth. Its standardized effect estimate was larger in magnitude than those of drought or fire.

    Bayesian Monod-curve model on tropical natural regrowth resurvey records (2014 records). Standardized effect on carbon accumulation annualized over the first 30 years: disruption −0.38 (95% CI −0.46 to −0.30); drought −0.19 (−0.27 to −0.10); fire −0.10 (−0.19 to −0.01); grazers −0.042 (−0.123 to 0.044, not significant). Areas with the lowest disruption accumulate four times as much carbon as the most disrupted. · Impacts on carbon accumulation; Forest carbon data and covariates; Forest carbon accumulation model; Tables

  2. SuggestedAt tropical-forest sites flagged as potential restoration areas, current disruption is estimated to cut yearly carbon accumulation potential by more than half.

    Mapped accumulation with disruption set to zero was compared against current (c. 2020) disruption, within restoration areas from Griscom et al. Mean loss was −1.8 Mg/ha/year, a 57% reduction for a typical site. The authors call this a maximum loss relative to the lowest observed disruption. · Assessing impacts on regrowth potential; Mapping estimates and uncertainty

  3. ShownThe negative link between disruption and carbon accumulation did not appear in monoculture plantations, where people planted the trees.

    The same model had an interaction term separating 1351 tropical monoculture plantation records from natural regrowth records. The disruption effect in plantations was 0.026 (95% CI 0.011 to 0.050), not negative. Plantation sites covered similar environmental ranges but were more common at higher latitudes and in cooler, drier places. · Impacts on carbon accumulation; Forest carbon data and covariates; Forest carbon accumulation model

  4. ShownMost tropical trees depend on animals for seed dispersal, and dependence rises toward the equator.

    The authors used 17,071 vegetation plots, with species weighted by abundance and by animal-dispersal dependence (from 346,501 dispersal-mode records). A GAM of prevalence against latitude gave F = 1893, P < 0.0001. A random forest map reached R^2 = 0.66 on a held-out test set (N = 4,258). On average, 81% of trees in tropical plots are animal-dispersed. · Dependence on animal seed dispersal; Prevalence of animal seed dispersal

  5. SuggestedTree cover change from 2000 to 2020 mostly increased disruption, so modeled natural regrowth potential fell in most places where it changed.

    Regrowth potential was mapped using disruption for c. 2000 vs c. 2020, updated for tree cover change. Among areas where potential changed by more than ±5%, 94% showed declines. The authors consider this conservative because many tree-cover gains are plantations. · Assessing impacts on regrowth potential; Calculating seed dispersal disruption; Mapping estimates and uncertainty

Methods, data and reproducibility

Published trait-based seed dispersal estimates for birds and mammals were adjusted for fragmentation effects on presence and human-footprint effects on movement, giving a ∼1 km tropical disruption map (proportional loss of dispersal function). A Bayesian Monod model of aboveground carbon over stand age, with disruption and environmental covariates, was fitted and used to map regrowth potential.

  • sPlotOpen woody vegetation plots (forest, natural landscapes), with SALVIAS plots taken from BIEN, giving species relative abundance · woody plants · n = 17,071 plots
  • Seed dispersal mode records compiled from plant trait databases and published studies, recoded as biotic vs abiotic · plants · n = 346,501 records; 24,660 species (24,600 in main text)
  • Published species-level seed dispersal function estimates (Fricke et al.) trained on dispersal networks, retention/movement data and gut-passage germination studies · birds and mammals · n = 406 networks; 302 animal species; 2215 studies
  • Meta-analysis dataset of local species occurrence in fragmented forest landscapes · birds (dataset focus) · n = 61,716 occurrence records; 1990 species
  • MoveBank animal tracking data under CC0 licences (Table S2) · terrestrial mammals and birds · n = 488,583 geolocations; 1838 individuals of 80 species; 7769 dispersal periods
  • Tropical natural forest regrowth aboveground biomass/carbon field records over stand age (Cook-Patton et al.) · n = 2014 natural regrowth records
  • Tropical monoculture plantation aboveground carbon records (Bukoski et al.) · n = 1351 records
  • Gridded layers: IUCN/BirdLife ranges, human footprint, tree cover (2000, 2020), WorldClim 2.1, VHI drought, Fire-cci v5.1, GLW 3 livestock, MODIS NDVI/NPP, Griscom et al. restoration map
Methods & tools
JAGS (rjags), mgcv, ranger, tidymodels, spatialsample, Google Earth Engine (rgee), BIOMASS R package, BIEN (rbien), Taxonomic Name Resolution Service, sPlotOpen, MoveBank, WorldClim 2.1
Reproducibility gaps
The text mentions a data and code package with full model details but gives no URL or accession. R package versions are not stated.
Limitations (6)
  • The precision of the conclusions depends on how much animal biodiversity data exists. Spatiotemporal mismatches between animal data and vegetation dynamics add uncertainty and bias, and between-species differences in disperser roles and responses to human drivers are not fully resolved. · stated by the authors
  • More drivers could be considered: expanded data may resolve effects of competing vegetation and land-use-history degradation (land-use history did not predict growth potential here). Pollination and herbivory also change under the same drivers; the authors argue this makes their estimates conservative. · stated by the authors
  • The field measurements come from a sample of tropical areas, so the results may carry spatial bias and miss biogeographic differences in how sensitive regrowth is to disruption. · stated by the authors
  • The restoration-site losses are a maximum relative to the lowest observed disruption. Only aboveground carbon is estimated; belowground and soil carbon are excluded. · stated by the authors
  • The design is observational and compares sites. Disruption is derived from fragmentation and human footprint, which also track other degradation; the covariates (drought, fire, grazers, potential NPP) and the plantation contrast reduce confounding but cannot rule it out. · noted in this summary
  • The relationship between species presence and fragmentation was fitted on a bird-focused occurrence dataset but also applied to mammals. The justification is a citation to similar mammal patterns, not mammal data analyzed in the paper. · noted in this summary

Read this if you model tropical forest carbon or restoration potential and want a pantropical, field-data test of how losing bird and mammal seed dispersers changes natural regrowth rates.

Citations
32
FWCI
26.42
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
THISLandim '25Hatfield '23Veryard '22Schnabel '25Miranda '25Cerullo '26
72% semantic matchEcology
Anna Rebello Landim, Jörg Albrecht, Jorge Brito+7 more·Current Biology·Sep 25, 2025· 9 citations