CoChat

Habitat fragmentation and its lasting impact on Earth’s ecosystems

Open access

The DOI matches its Crossref record; 123 of the 123 references it cites were found in a registry or an index; 12 of the 85 sources its claims cite were checked for support.

Abstract

Urgent need for conservation and restoration measures to improve landscape connectivity.

Overview

CoChat Summary

TL;DR

Long-term fragmentation experiments spanning 35 years and five continents show that habitat fragmentation consistently degrades biodiversity and ecosystem functions, with effects that often strengthen over time. Global forest analysis finds most remaining forest is near an edge, and the experimental synthesis supports extinction debt, immigration lag, and ecosystem function debt.

Key findings

  • Nearly 20% of the world’s remaining forest is within 100 m of an edge, and more than 70% is within 1 km. In the Brazilian Amazon, forest farther than 1 km from the edge decreased from 90% historically to 75% today; in the Brazilian Atlantic Forest, it fell from 90% to less than 9%. Also, 37% of the world’s grassland eco-regions are classified as “highly fragmented.”
  • Across 76 studies in the five longest-running experiments, reduced fragment area, increased isolation, and increased edge reduced species persistence, species richness, nutrient retention, trophic dynamics, and movement. Species richness of plants, arthropods, and birds in mature forest fragments and replicated moss landscapes decreased by 20 to 75% after fragmentation.
  • Temporal lags were common: average species loss was >20% after 1 year and >50% after 10 years, still increasing in the longest time series (>2 decades). Immigration lags resulted in 5% fewer species after 1 year and 15% fewer after 10 years in small or isolated fragments; ecosystem function debt amounted to 30% loss after 1 year, rising to 80% after a decade.
  • Larger fragments paid extinction debt more slowly: in BDFFP, a 50% decline in bird species occurred after 5 years in 1-ha fragments but after 12 years in 100-ha fragments; in Moss, arthropod species richness declined 40% in small fragments and 26% in large fragments after 1 year.

Methods

  • Global analysis used the world’s first high-resolution map of global tree cover to measure distance to forest edge, plus historical reconstruction of forest extent and fragmentation in the Brazilian Amazon and Brazilian Atlantic Forest.
  • Synthesis covered the entire set of ongoing terrestrial long-term fragmentation experiments, spanning 35 years and disparate biomes on five continents; it included 76 studies across the five longest-running experiments, with results available 31 January 2014.
  • Fragmentation effects were calculated as log response ratios for three treatments—reduced fragment area, increased fragment isolation, and increased proportion of edge—compared with non- or less-fragmented habitats; time-series effects were estimated with linear mixed random-slopes models.

Limitations

  • The paper states that conclusions from experiments thus far are likely conservative because impacts from other environmental changes have been mostly excluded, and that current experiments have stopped short of examining evolutionary responses such as genetic bottlenecks, ecological traps, changing selection, inbreeding, drift, and gene flow.
  • The paper states that the global forest analysis underestimates the total number of fragments in the smallest bin (1 to 10 ha) and that a similar spatial analysis is not currently possible in grasslands.
  • Evident from the text: the synthesis is restricted to terrestrial long-term experiments and to published and unpublished studies available by 31 January 2014; effect sizes are statistical, so positive or negative values do not uniformly mean degrading function for every species or process.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1126/sciadv.1500052

Cites

123 of 123 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 The global magnitude of forest fragmentation. (A) Mean distance to forest edge for forested pixels within each 1-km cell. Lines point to locations of ongoing fragmentation experiments identified and described in Fig. 2. (B) Proportion of the world’s forest at each distance to the forest edge and the cumulative proportion across increasing distance categories (green line). (C and E) In the Brazilian Amazon (C) and Atlantic Forests (E), the proportion of forest area at each distance to forest edge for both the current and estimated historic extent of forest. (D and F) In the Brazilian Amazon (D) and Atlantic Forests (F), the number of fragments and the total area of fragments of that size. The total number of fragments in the smallest bin (1 to 10 ha) is an underestimate in both the Atlantic Forest and Amazon data sets because not all of the very smallest fragments are mapped.
Fig. 1 The global magnitude of forest fragmentation. (A) Mean distance to forest edge for forested pixels within each 1-km cell. Lines point to locations of ongoing fragmentation experiments identified and described in Fig. 2. (B) Proportion of the world’s forest at each distance to the forest edge and the cumulative proportion across increasing distance categories (green line). (C and E) In the Brazilian Amazon (C) and Atlantic Forests (E), the proportion of forest area at each distance to forest edge for both the current and estimated historic extent of forest. (D and F) In the Brazilian Amazon (D) and Atlantic Forests (F), the number of fragments and the total area of fragments of that size. The total number of fragments in the smallest bin (1 to 10 ha) is an underestimate in both the Atlantic Forest and Amazon data sets because not all of the very smallest fragments are mapped.CC BY · PubMed Central
Fig. 2 The world’s ongoing fragmentation experiments. All experiments have been running continuously since the time indicated by the start of the associated arrow (with the exception of the moss fragmentation experiment, which represents a series of studies over nearly two decades). The variables under study in each experiment are checked. The area is that of the experiment’s largest fragments. Icons under “Fragment” and “Matrix” indicate the dominant community and its relative height, with multiple trees representing succession.
Fig. 2 The world’s ongoing fragmentation experiments. All experiments have been running continuously since the time indicated by the start of the associated arrow (with the exception of the moss fragmentation experiment, which represents a series of studies over nearly two decades). The variables under study in each experiment are checked. The area is that of the experiment’s largest fragments. Icons under “Fragment” and “Matrix” indicate the dominant community and its relative height, with multiple trees representing succession.CC BY · PubMed Central
Fig. 3 Fragmentation effects propagate through the whole ecosystem. (A to C) For each fragmentation treatment [reduced area in BDFFP, Wog Wog, Kansas (A); increased isolation in SRS and Moss (B); and increased edge in all experiments (C)], we summarize major findings for ecological processes at all levels of ecological organization. Each dot represents the mean effect size [computed as log response ratio: ln(mean in more fragmented treatment/mean in non- or less-fragmented treatment)] for an ecological process. Effect sizes are statistical, such that negative or positive values could represent degrading function. Horizontal bars are the range when a dot is represented by more than one study. Details, including individual effect sizes for each study, are reported in table S1.
Fig. 3 Fragmentation effects propagate through the whole ecosystem. (A to C) For each fragmentation treatment [reduced area in BDFFP, Wog Wog, Kansas (A); increased isolation in SRS and Moss (B); and increased edge in all experiments (C)], we summarize major findings for ecological processes at all levels of ecological organization. Each dot represents the mean effect size [computed as log response ratio: ln(mean in more fragmented treatment/mean in non- or less-fragmented treatment)] for an ecological process. Effect sizes are statistical, such that negative or positive values could represent degrading function. Horizontal bars are the range when a dot is represented by more than one study. Details, including individual effect sizes for each study, are reported in table S1.CC BY · PubMed Central
Fig. 4 Delayed effects of fragmentation on ecosystem degradation. (A) The extinction debt represents a delayed loss of species due to fragmentation. (B) The immigration lag represents differences in species richness caused by smaller fragment area or increased isolation during fragment succession. (C) The ecosystem function debt represents delayed changes in ecosystem function due to reduced fragment size or increased isolation. Percent loss is calculated as proportional change in fragmented treatments [for example, (no. of species in fragment − no. of species in control)/(no. of species in control) × 100]. Fragments and controls were either the same area before and after fragmentation, fragments compared to unfragmented controls, or small compared to large fragments. Filled symbols indicate times when fragmentation effects became significant, as determined by the original studies (see table S2). Mean slopes (dashed lines) were estimated using linear mixed (random slopes) models. Mean slope estimates (mean and SE) were as follows: (A) −0.22935 (0.07529); (B) −0.06519 (0.03495); (C) −0.38568 (0.16010).
Fig. 4 Delayed effects of fragmentation on ecosystem degradation. (A) The extinction debt represents a delayed loss of species due to fragmentation. (B) The immigration lag represents differences in species richness caused by smaller fragment area or increased isolation during fragment succession. (C) The ecosystem function debt represents delayed changes in ecosystem function due to reduced fragment size or increased isolation. Percent loss is calculated as proportional change in fragmented treatments [for example, (no. of species in fragment − no. of species in control)/(no. of species in control) × 100]. Fragments and controls were either the same area before and after fragmentation, fragments compared to unfragmented controls, or small compared to large fragments. Filled symbols indicate times when fragmentation effects became significant, as determined by the original studies (see table S2). Mean slopes (dashed lines) were estimated using linear mixed (random slopes) models. Mean slope estimates (mean and SE) were as follows: (A) −0.22935 (0.07529); (B) −0.06519 (0.03495); (C) −0.38568 (0.16010).CC BY · PubMed Central

Verification

Claims the paper makes

7 Valid · 2 Not valid · 40 Not checked · 57 Own results and interpretations

66 more claims · 2 with their cited sources checked

Create a free account to see all 106 claims and to verify the 73 references still open.

See all claims — free account
  • It may state two separate claims as one. 30
  • Probably not a substantial claim. 7

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

2 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Citations

Cites

Show all 123 references

Research works printed but not linked: 2

  1. T. Ricketts, E. Dinerstein, D. Olson, C. Loucks, W. Eichbaum, K. Kavanagh, P. Hedao, P. Hurley, K. Carney, R. Abell, S. Walters, A Conservation Assessment of the Terrestrial Ecoregions of North America (Island Press, Washington, DC, 1999).
  2. E. O. Wilson, E. O. Willis, in Ecology and Evolution of Communities, M. L. Cody, J. M. Diamond, Eds. (The Belknap Press, Cambridge, MA, 1975), pp. 522–534.

Cited by

Items 301–400 of 1,005

Metadata

Authors
Haddad, N. M., Brudvig, L. A., Clobert, J., Davies, K. F., Gonzalez, A., Holt, R. D., Lovejoy, T. E., Sexton, J. O., Austin, M. P., Collins, C. D., Cook, W. M., Damschen, E. I., Ewers, R. M., Foster, B. L., Jenkins, C. N., King, A. J., Laurance, W. F., Levey, D. J., Margules, C. R., Melbourne, B. A., Nicholls, A. O., Orrock, J. L., Song, D.-X., & Townshend, J. R.
Affiliations
  • North Carolina State University (US)
  • Michigan State University (US)
  • Centre National de la Recherche Scientifique (FR)
  • University of Colorado Boulder (US)
  • McGill University (CA)
  • University of Florida (US)
  • Department of Environmental Science and Policy, George Mason University, Fairfax, VA 22030, USA.
  • Global Land Cover Facility, Department of Geographical Sciences, University of Maryland, College Park, MD 20702, USA.
  • Commonwealth Scientific and Industrial Research Organisation (AU)
  • Colby College (US)
  • St. Cloud State University (US)
  • University of Wisconsin–Madison (US)
  • Imperial College London (GB)
  • University of Kansas (US)
  • Instituto de Pesquisas Ecológicas (BR)
  • James Cook University (AU)
  • U.S. National Science Foundation (US)
  • University of Indonesia (ID)
  • University of Maryland, College Park (US)
Corresponding author
Haddad, Nick M.
Venue
Science Advances
Published
2015-03-06
Volume
1
Issue
2
Pages
e1500052-e1500052
ISSN
2375-2548 (ISSN-L)
OpenAlex
W2127597920
Cited by
4,773
References
123
Citations by year
2026: 672 · 2025: 622 · 2024: 537 · 2023: 491 · 2022: 525 · 2021: 445 · 2020: 412 · 2019: 376 · 2018: 282 · 2017: 215 · 2016: 146 · 2015: 41
FWCI
237.74
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • Science Advances (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • ResearchOnline at James Cook University (James Cook University) (submittedVersion)
  • Florida International University Digital Commons (Florida International University) (submittedVersion)
  • Florida International University Digital Commons (Florida International University) (submittedVersion, cc-by)
  • HAL (submittedVersion, other-oa)
  • eScholarship@McGill (McGill) (submittedVersion)
  • KU ScholarWorks (The University of Kansas) (submittedVersion, cc-by)
Funders
  • National Science Foundation (0108302, 0841892, 0950100)
  • National Aeronautics and Space Administration (NNH07ZDA001N-LCLUC, NNH07ZDA001N, NNH06ZDA001N)
  • U.S. Department of Agriculture
  • Andrew W. Mellon Foundation
  • University of Florida Foundation
  • Laboratoire d'Excellence TULIP
  • Yayasan Sime Darby
  • Commonwealth Scientific and Industrial Research Organisation
  • Agence Nationale de la Recherche
  • Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
  • Analyses et Expérimentations pour les Ecosystèmes
  • U.S. Forest Service
  • Ciência sem Fronteiras
  • FP7 Ideas: European Research Council (281986)
Topics
  • Physical Sciences › Environmental Science › Nature and Landscape Conservation › Ecology and Vegetation Dynamics Studies
  • Physical Sciences › Environmental Science › Ecology › Wildlife-Road Interactions and Conservation
  • Physical Sciences › Environmental Science › Global and Planetary Change › Conservation, Biodiversity, and Resource Management
Keywords
habitat fragmentation, forest cover, forest edge, biodiversity loss, ecosystem functioning, nutrient cycling, biomass, fragment size, isolation, landscape connectivity, conservation, ecosystem services, extinction

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.